💬│general

[storm kindle]

its an extension of my brain atp

[storm kindle]

i dont feel complete without it

[timber niche]

Sam is a bot, confirmed

Can't do shit

[limpid flare]

Does anyone know if there is like a feedback section in this server?

[storm kindle]

#👉│feedback-general #💻︱feedback-desktop #📱︱feedback-mobile

[limpid flare]

Thank you

[distant aurora]

@storm kindle https://cunnyx.com/LittleB_inV/status/1928967552455115042

LittleB (@LittleB_inV)

New martyrdom buff makes it so grenades explode before you can even finish an execution animation.😂😂😂 #CODm #CoDMobile

💬 16 🔁 2 ❤️ 103 👁️ 2.9K

▶ Play video

CunnyX · Jun 1, 2025, 12:11 AM

Best buff ever

[storm kindle]

i heard

they buffed it to 1s

[tame current]

https://youtu.be/7ueKBY56YcE?si=lQcK6lJNgvak3Obo

check this out

[short narwhal]

Input rates

[distant aurora]

But thinking is unpredictable.

[short narwhal]

That’s why it gets costly

[distant aurora]

So if it gets stuck in a loop I'll have a fat bill.

💀

[short narwhal]

There’s token limits for how much a model can think for that helps with that

Etc on Gemini and Claude

[tranquil orchid]

Think you found the weirdest AI response?
Prove it—and score a hoodie!

Wrongest Answer Contest — Now Live!

Full announcement & prizes:
#📰│server-news message

**How to enter:**

• Drop a screenshot + thread link in
  https://discord.com/channels/1047197230748151888/1373806851358986320
• Vote for your faves with
  Submit by June 1, vote by June 8.

You could win a Pro sub + merch!

[distant aurora]

Does open router have it?

[short narwhal]

🤷

[distant aurora]

Couldn't see anything in sampling parameters.

[short narwhal]

Prolly not then?

Depends on the model too I guess

[distant aurora]

That's o4 mini.

[short narwhal]

I haven’t used o4 mini through the api ever so not sure

It has the low, medium high thinking modes that are basically that

[gloomy finch]

Maybe the max tokens one do the job of limiting it

[distant aurora]

@left stream

[distant aurora]

I thought that's for output.

[short narwhal]

Is there reasoning effort anywhere?

[distant aurora]

Nope

That's all there is.

[distant aurora]

High and medium are as separate models instead.

[short narwhal]

Straight from openrouters docs

[distant aurora]

Wait. So if that controls thinking limit, what about the response limit?

[short narwhal]

😬not sure

[distant aurora]

Anyways, thanks for responding.

https://tenor.com/view/keyaan-john-wick-keanu-reeves-gif-2384046765437701795

[modern sinew]

[plush canyon]

What are the usage limits for Perplexity Labs?

[short narwhal]

Wait no I’m stupid they’re charged at output rates 😭

[short narwhal]

50 uses a month

[distant aurora]

This confused me.

https://chatgpt.com/share/683c1ea4-2854-8013-a2d5-f1a2d689830b

[short narwhal]

Yeah the 2. Step

[distant aurora]

"I have asked to include"

[short narwhal]

Ignore that

[distant aurora]

By that logic I should have less costs if I ask it for a one word response to a complex math problem.

[short narwhal]

Like 10 tokens less prob

It still needs to reason to figure out the problem

[distant aurora]

@short narwhal I don't feel very positive about the deep research.

Gpt 4o (o4 mini high)

@vague pagoda When I use an ai model via api, is thinking charged at input or output rates? Or is it not charged at all? I'm asking if I will be charged for the lengthy thinking if I ask it for a one word/like response to a complex math problem that requires hundreds of lines for reasoning

[short narwhal]

Don’t know what it’s on about

100% charges at output rate

[timber niche]

Wonder how the compute is factored in then

Because thinking can consume alot of compute if it's the scenario billy stated

[short narwhal]

True, but then that’s where reasoning effort comes into play

low o4-mini reasons less

But it is pricy still for sure

[timber niche]

What is reasoning effort quantified by?

[distant aurora]

Some unknown token limit for thinking.

[timber niche]

Why would they keep it unknown

Makes no sense

But hey if it's not gonna cost you, we take the W

[foggy stone]

hi

[short narwhal]

👋

[distant aurora]

If google is 25000

Open ai should be pretty low.

Even qwen 3 has 36000 I think.

[gloomy finch]

Hello

[spring dove]

Helloooo

[timber niche]

Hola

[distant aurora]

https://tenor.com/view/morning-star-hello-there-fallen-angel-devil-lucifer-gif-12199746

[gloomy finch]

O3 says there's 256,
https://you.com/search?q=in+here+through+Google+ai+studio+it+shows+there+is+around+24576+as+thinking+token+limit+then+y+is+open+ai+thinking+limit+so+low%3F&cid=c1_798cf8cb-3915-4360-8fc5-526f5a67a882&tbm=youchat

[left stream]

Openai has it confirmed that certain models costs them more money than it’s making them

[tranquil orchid]

Think you found the weirdest AI response?
Prove it—and score a hoodie!

Wrongest Answer Contest — Now Live!

Full announcement & prizes:
#📰│server-news message

**How to enter:**

• Drop a screenshot + thread link in
  https://discord.com/channels/1047197230748151888/1373806851358986320
• Vote for your faves with
  Submit by June 1, vote by June 8.

You could win a Pro sub + merch!

[left stream]

Even with openai pro plan

But it’s a no brainer considering models like o1 pro are unlimited like gpt 4.5 and o3

[distant aurora]

Answer this: When I use an ai model via api, is thinking charged at input or output rates? Or is it not charged at all? I'm asking if I will be charged for the lengthy thinking if I ask it for a one word/like response to a complex math problem that requires hundreds of lines for reasoning

[distant aurora]

256 tokens is too less. That can't be true.

Thinking tokens can't be that less than response tokens afaik.

Even the thinking for my math problem exceeds 256 tokens I think.

[frosty vessel]

hi

[tame current]

ola

[left stream]

“output tokens include internal reasoning tokens generated by the model that are not visible in API responses.”

https://community.openai.com/t/how-to-correct-compute-the-cost-of-an-o1-model-api-call/1099176?

[distant aurora]

Thanks 🙏🫶

[timber niche]

Good info

[distant aurora]

Something as simple as this is 816 tokens. The response alone. https://you.com/search?q=There's+a+square+edfb+of+6cm+inside+a+right+angled+triangle+abc+(b+being+the+right+angle)+with+hypotenuse+ac+%3D+20cm+passing+through+d.+Ab+passes+through+e.+Find+length+of+ab&cid=c1_839f2d18-3a87-44dc-84b2-d5d1dd53a866&tbm=youchat

Thinking would be wayyy longer.

[left stream]

Yes

[distant aurora]

Also, the O3 on you.com is kinda dumbed down.

[left stream]

Ngl i think you have a point

I’ll test it in you.com later once more

[gloomy finch]

Yup

[left stream]

I forgot to reply to you

When you said the jb didn’t work

Was personalization turned off in you.com?

That messes up with the jb too sometimes. Also, ensure to disable search

I forgot the third thing you had to disable since I’m not on you.com like that

[distant aurora]

Is it on or off by default?

[left stream]

On by default I think

Should be off

[left stream]

Disable the second option that comes included in the tab where personalization is at

[atomic shell]

trying labs mode of perplexity the timing is kinda bugged

it was 5 minute snot it says 7

[timber niche]

Happens

Happened to me

[atomic shell]

yes

[gloomy finch]

U mean both options ?

Or just the smart learn

[left stream]

Yes disable both

[gloomy finch]

Alright

[lone hound]

Well, it is logical, it has several unnecessary models available, instead of having 1 or 2 models in operation. The owner of the company does not think about it.

🧠

[sharp root]

what are the top ai news today

[left stream]

True, the ones they could keep is o1 pro, o3, 4o, and 4.5 (arguable)

[lone hound]

It's very simple: competition between companies and they are tools in my hands.

🙂

[lone hound]

Why 4o and not 4.1?

[distant aurora]

You mean 4.1 instead of 4o?

[distant aurora]

I know a telegram channel that can keep you reasonably updated.

Ah.

@tame current @pseudo ferry https://www.instagramez.com/p/DKWIScQStb0

The Trending Indian 🇮🇳 (@thetrendingindian)

💬 23 🔁 0 💜 3.1K 👀 0

India’s active COVID-19 cases have climbed to 3,395, with Kerala reporting the highest at 1,336. Maharashtra and Delhi follow with 467 and 375 cases. Four d*aths were recorded in the last 24 hours. Health officials link the surge to new Omicron sub-variants NB.1.8.1 and LF.7, though most cases are mild. Citizens are urged to stay alert, not panic.

Follow @thetrendingindian for more latest News updates 🗞️

#India #News #Fyp #Explore

EmbedEZ • 1 - 1 of 1 · Jun 1, 2025, 5:04 AM

[timber niche]

Dang.

[tame current]

https://tenor.com/view/shock-shocker-shocked-black-guy-gif-27526370

[timber niche]

Guess i know where not to travel

In 2025, that's a huge number

[left stream]

Hes gonna go bye bye

[distant aurora]

Fr

[left stream]

I forgot you can search with 4.1, it has a million context memory right?

[distant aurora]

You.com is capped at 64k

Also, doesn't matter.

[left stream]

Really? Crazy

[distant aurora]

Also, you have to do this on you.com everytime just to get quality search #💬│general message

Or else it can't even tell the lyrics of a song.

@left stream is this real? https://youtube.com/shorts/UsQCkoGRVMY

[spring sundial]

hey - where is the best channel to ask some techncial questuons about Perplexity - not really support ones

[valid charm]

Here

[valid charm]

Yeah and their deep research thingy is the worst

[atomic shell]

oooo

[royal ingot]

Has there been any issues with perplexity support lately? because it has been over 3 days and they haven't responded. i have sent 2 emails and also sent messages via the in-built support chat but still no response on both platforms

[tranquil orchid]

Think you found the weirdest AI response?
Prove it—and score a hoodie!

Wrongest Answer Contest — Now Live!

Full announcement & prizes:
#📰│server-news message

**How to enter:**

• Drop a screenshot + thread link in
  https://discord.com/channels/1047197230748151888/1373806851358986320
• Vote for your faves with
  Submit by June 1, vote by June 8.

You could win a Pro sub + merch!

[tame current]

yeah, many people complained that they are responding late

[distant aurora]

Fr

[royal ingot]

hmmm, would you be able to what is the approx response time?

[tame current]

nope

[distant aurora]

@left stream interesting thread https://cunnyx.com/immasiddx/status/1928821279500484960

sid (@immasiddx)

Dear Apple,
︀︀
︀︀You had ONE job. 😭

💬 125 🔁 43 ❤️ 960 👁️ 80.1K

CunnyX · May 31, 2025, 2:30 PM

https://cunnyx.com/Sankew06/status/1928824645504864712

Sankew (@Sankew06)

@immasiddx bro😭

💬 12 🔁 2 ❤️ 91 👁️ 9.7K

CunnyX · May 31, 2025, 2:43 PM

[timber niche]

Lmao

[royal ingot]

right. i have seen people not recieving a response at all even after 3 weeks or smth. they haven't sorted that out yet?

i mean at least for pro users the response should be prioritised

[tame current]

yeah but it is what it is

[timber niche]

Sam responds to me well tho

[royal ingot]

sad

[royal ingot]

yeah me too. its its "teammate" which he forwarded the request to, is taking too much time 😅

[timber niche]

LOL

Human vs AI

[rapid goblet]

Random Chinese writing and links in the responses - anyone?

[valid charm]

Ask him about Cursor comming out?

[timber niche]

Cursor is removed not by pplx but by cursor themselves

They have found too many people abusing it

[spring sundial]

Thanks - I'm doing a lot of AI research and learning at the moment - and find ChatGPT by itself limiting. I do a lot of image work as well - and I was interested particularly by the model agnostic featires of Perplexity. So I signed up for a month Pro to try it... What I am confused / annoyed about so far - is that you don't appear to be able to choose which models imaging engine to use - which was a big factor for me - as I use both ChatGPT 4o and Gemini pro for imaging - and interested to use as many variants as possible - eg Grok or Anthropic for example. I am also try ing to see how I can replicate my Custom GPT's successfully - in terms of loading in files and knowledge and having them behind the scenes so user's can't see the resources, but only the responses.

[distant aurora]

Send him this.

Gemini passes it btw.

[left stream]

Claude opus 4 passed the image gas bill test

[distant aurora]

Native?

[left stream]

Yes

[distant aurora]

I'm free user so I don't have thinking mode.

Non thinking sonnet fails

[left stream]

This is sonnet 4 thinking

Looking at this Maharashtra State Electricity (MSEDCL) bill, I'll explain the key points in simple terms:

Current Bill Details:

• Bill Period: May 2025
• Due Date: June 9, 2025
• Current Month Usage: Only 1 unit
• Amount Due: Rs. 40.00

Electricity Usage for Last 6 Months:

Based on the consumption graph shown in the middle of the bill, here's your usage from December 2024 to May 2025:

1. December 2024: ~110 units
2. January 2025: ~125 units
3. February 2025: ~110 units
4. March 2025: ~70 units
5. April 2025: ~115 units
6. May 2025: 1 unit (current month)

Key Observations:

• Your consumption has been quite low in May 2025 (only 1 unit), which is unusual compared to previous months
• Your average consumption over the past 6 months (excluding May) was around 106 units per month
• The lowest consumption before May was in March 2025 (70 units)
• The highest was in January 2025 (125 units)

The bill shows a minimum charge of Rs. 40 despite very low consumption, which is likely the fixed/minimum charge applied by MSEDCL even when usage is minimal.​​​​​​​​​​​​​​​​

[distant aurora]

Wrong again.

[left stream]

Oof sonnet 4 thinking and opus 4 thinking gap is crazy

[distant aurora]

Show opus result.

[left stream]

Looking at this MSEDCL bill more carefully, here's what it actually shows:

This Month's Summary:

• Meter Reading Date: May 16, 2025
• Previous Reading: 20,069 kWh (April 15, 2025)
• Current Reading: 20,069 kWh (May 16, 2025)
• Electricity Used: 0 units - your meter didn't move at all!
• Amount to Pay: ₹40.00 (fixed minimum charge)
• Payment Due: June 9, 2025

Why ₹40 When You Used Nothing?

Even though you consumed zero electricity this month, MSEDCL charges a standard ₹40 minimum service fee. This fixed charge keeps your connection active regardless of usage. Since both meter readings are identical at 20,069 kWh, you're only paying this basic connection fee.

Your 6-Month Usage Pattern:

Looking at the consumption graph on the bill:

• December 2024: 81 kWh
• January 2025: 94 kWh
• February 2025: 90 kWh
• March 2025: 105 kWh
• April 2025: 144 kWh (highest usage)
• May 2025: 0 kWh (current month)

Your consumption typically ranges between 80-145 units per month. April saw your peak usage at 144 units, while December was the lowest at 81 units (excluding this month's zero). The sudden drop to zero in May is quite unusual compared to your regular pattern.

To pay, simply scan the QR code with any UPI/BHIM app for quick payment.​​​​​​​​​​​​​​​​

@distant aurora what do you think

[timber niche]

Quite the difference

[distant aurora]

Accurate

So Opus has quality image analysis but sonnet doesn't.

O4 mini has it.

Gemini has it on ai studio and third party but not natively, which is weird.

[timber niche]

Did you try through ai studio yet

[distant aurora]

Yes

[timber niche]

But you weren't happy with it right

[distant aurora]

I was.

[timber niche]

Hm

I thought only gpt aced it

[craggy geode]

guys i need help, gemini 2.5 pro gave me "79", claude sonnet 4 gave me "50", perplexity gave me "41"

https://www.perplexity.ai/search/if-i-have-150-pieces-and-each-w2F5xpw_T2m5PkrdGgLFaQ

prompt: "if i have 150 pieces, and each piece is a repeated 0 to 9.

then as far as the combinations go, like for instance how it would take a 2 and 8 to make 28, what's the highest number i could go to

not how many digits, like the combination if i incremented starting at 1, then 2, 3, so on, 12, 14, see what i mean? those 12, and 14, both took a 1

factor in the fact that like 21, 22, 23, 24, all use a 2, same with 3 and 31 32 33 etc"

[distant aurora]

Why did you not try o4 mini?

[craggy geode]

where try?

[distant aurora]

Or R1?

In perplexity itself.

[craggy geode]

i'm refining the prompt

[green crystal]

i got 79

[left stream]

I use opus only with o1 pro for coding. Main reason I got it

[green crystal]

o4 mini high just used python lol

[craggy geode]

i think this is actually a difficult counting problem, or am i actually just being stupid?

[left stream]

@distant aurora

[distant aurora]

Pac-Man

Can you do the same in chatgpt too?

[left stream]

No clue, haven’t read on what codex can do to the max

[craggy geode]

sonnet 4 gives "50", gemini 2.5 pro gives "50", perplexity gives "41"

[distant aurora]

I got 50

Twice

On both t3 chat and you.com

[green crystal]

lemme try o3 lol

[craggy geode]

DAMN i got Gemini 2.5 Pro to think for a solid 3 minutes on that one wtf

[timber niche]

How do you access o1 pro?

[left stream]

I’ll test it with the big guns

[distant aurora]

200 dollars per month

[left stream]

Pro subscription in chatgpt

[timber niche]

I looked it up and it said only through api

[left stream]

That too

[craggy geode]

i'm starting to think this is correct, but that's just purely based on consistent response with this refined prompt:

"If I have 150 digit tiles, with exactly 15 of each digit from 0 to 9, and I use these tiles to build consecutive integers starting from 1 (using up each digit as it appears in the numbers), what is the highest integer I can construct before running out of any digit?"

[timber niche]

Joe everyday

[green crystal]

yeah o3 hit me with 50

[craggy geode]

oh used my OG prompt nice thanks

[green crystal]

i thought o4 mini was supposed to be slightly better at math

[distant aurora]

I used o4 mini too.

Lol.

You sure you used high?

[green crystal]

yep
but it thought very short because it figured it could just use python

[distant aurora]

I'mma try it with gawk.

[craggy geode]

i did not expect this to be some crazy counting problem 😂

[timber niche]

Tbh this kinda problems usually broke models before

[left stream]

Because you only have 15 copies of each digit 0-9 (150 pieces total), every time a digit appears in the running count you permanently “spend” one copy of that digit.

⸻

Step-by-step count of how fast the digits run out

Block of numbers written Extra copies of each tens-digit used Cumulative copies of the most-used digits after that block
1 – 9 – 1 copy of each 1-9
10 – 19 ten 1’s 1’s → 12 copies
20 – 29 ten 2’s 2’s → 13 copies
30 – 39 ten 3’s 3’s → 14 copies
40 – 49 ten 4’s 1’s, 2’s, 3’s, 4’s → 15 copies each (limit reached)

At the end of 49 you have:
• 1’s, 2’s, 3’s, 4’s → all 15 copies used up
• 0’s → 4 copies used
• 5’s, 6’s, 7’s, 8’s, 9’s → 5 copies used each

⸻

Can you write 50?

Yes. 50 needs one more 5 and one more 0.
After writing 50 the tallies are:
• 5’s → 6 copies
• 0’s → 5 copies
• All other digits unchanged (still ≤ 15)

So 50 is still possible.

⸻

Why 51 (and everything beyond) is impossible

51 would need another 1, but every 1 is already gone (15/15 used).
Because you can’t reuse pieces, the count stops there.

⸻

Highest number you can reach

50

That is the last integer you can form while writing every whole number sequentially from 1 upward without exceeding the 15-copy limit on any digit.

[craggy geode]

that's very clean logic i like it, thanks

[left stream]

Here’s how the table would look

So it doesnt look confusing

[distant aurora]

https://x.com/i/grok/share/KMhP2XuiQoudRMyg37xeWs1VV

if i have 150 pieces, and each piece is a repeated 0 to 9. then as...

Let’s dive into this problem with a fresh perspective. You have 150 pieces, where each piece is a digit from 0 to 9, and these digits can repeat. The goal is to determine the highest number you can reach by writing numbers in sequence—starting at 1, then 2, 3, and so on (like 12, 14)—using up the digits from your 150 pieces as you go. You...

X

[craggy geode]

thanks for helping me out guys, that was an interesting one i think

[left stream]

Lol this one made all of them think

[timber niche]

Do you have a cap on o1 pro and opus?

[valid charm]

Where is this from??

[left stream]

Chatgpt

[timber niche]

Chat gpt pro

[left stream]

O1 pro is unlimited for me

Opus i have a cap

[timber niche]

Nice

How much

[left stream]

Tbh no clue I haven’t hit the limit yet

I heard it resets after 4 hours

For the $20 sub

[timber niche]

Then essentially unlimited

Wait 20 usd sub has opus

[left stream]

Yes

[timber niche]

How did i miss that that's a great deal then

[left stream]

Ikr api is crazy

It drains fast

[timber niche]

4 sonnet is already great

No glazing

Straight results

[left stream]

True, it’s impressive. Been using it on perplexity

[timber niche]

Why not directly on Claude

[left stream]

I use that too, I use it on both

[timber niche]

Ah

[left stream]

Perplexity is impressive for search ngl

[timber niche]

I for some reason don't like claude's ui

Something about the grey throws me off

Yes just looking up stuff it's best

[left stream]

True, perplexity’s ui looks better and I prefer using it on there.

[green crystal]

wonder why plus doesnt even get to try it
is o1 pro that much more expensive than o3?

[left stream]

Yes

[green crystal]

[left stream]

It’s the most expensive model openai has to offer

Hopefully it gets added to plus one day

[short narwhal]

600 per million output token 💔

[timber niche]

For normal users more advanced voice mode is better than having o1 pro

[green crystal]

more advanced mode?

[timber niche]

Voice*

[left stream]

I could buy a lot of vanilla cones with that damn

[green crystal]

erh guess its another currency?

[left stream]

Dollars

[short narwhal]

Dollars yeah

[green crystal]

$600 per million?

[timber niche]

That's just absurd

[left stream]

Yes

[green crystal]

[short narwhal]

For output

Input is 150?

[green crystal]

dafuq lol

[left stream]

Yes

Thats why pro users take advantage of that

O1 pro is unlimited on there

[timber niche]

No wonder they loose money on pro subs

[left stream]

Ikr

10000 messages on the first week when I first got it 😆

[green crystal]

friggin crazy pricing

[timber niche]

Damn you're gonna bankrupt them

Did you get it mainly because of work?

[left stream]

Partially

Studying too

[timber niche]

Nice

[left stream]

O1 pro was insanely impressive for comptia

[timber niche]

The base cert?

[left stream]

Yeah back then I used it for that and passed thanks to it

[timber niche]

Congratulations haha

[left stream]

Thanks

[timber niche]

No longer have to look for dumps

[left stream]

O3 would find them 🤣

[tame current]

WALLPAPER

[timber niche]

Wow

[timber niche]

That's cool

[wanton sluice]

what should the answer be anyways

[craggy geode]

50

[wanton sluice]

o4 mini high said 50

[craggy geode]

yea i think we reached consensus, it got the models thinking quite a bit

[wanton sluice]

even this cheap model

[left stream]

@distant aurora

[wanton sluice]

i know you got your answer i'm just testing models

it's interesting to me to know which can get it and which cannot

interestingly, claude opus on extended thinking mode is still iterating for like 2 minutes

[craggy geode]

you using my original prompt, or refined prompt?

duck ai doesn't want to answer

answere this immediately though

🤔

i think they put something in duck ai to make it not respond if it goes beyond a certain threshold of compute

anyone else experience that?

confirmed, duck ai has a threshold

alright i'm good on that one, 50 it is ☺️

[distant aurora]

Nice.

But does the effect stay with follow up prompts?

[timber niche]

Joe's phone prolly died

Lol

[left stream]

Wdym

[distant aurora]

Will it stay broken or do I have to paste the whole thing everytime I want an uncensored response?

[left stream]

It should stay broken

As long as its in the same thread

If its not broken

Start a new thread with the same prompt

[distant aurora]

https://www.instagramez.com/p/DKWbzftBQoT

[serene sand]

Shaping AI for the Public Good | Dr. Andy Konwinski and Prof. David Patterson ☕

[opal prairie]

What do you guys think about the labs in perplexity?

[tranquil orchid]

Think you found the weirdest AI response?
Prove it—and score a hoodie!

Wrongest Answer Contest — Now Live!

Full announcement & prizes:
#📰│server-news message

**How to enter:**

• Drop a screenshot + thread link in
  https://discord.com/channels/1047197230748151888/1373806851358986320
• Vote for your faves with
  Submit by June 1, vote by June 8.

You could win a Pro sub + merch!

[molten sparrow]

I am pretty amazed, as we live close to a river which tend to provoke flodings from time to time I asked Perplexity Labs to create a dashboard for monitoring the water levels between Passau and Budapest with a simple prompt: "Build a dynamic dashboard that can be accessed in a web browser, which allows observing the water levels of the Danube River between Passau and Budapest at at least 10 different measuring points. Current status in comparison to the normal water level and with additional hourly data collection, course, and graphical representation. Additionally, a display of the function and status of the locks along the way" and teh result was amazing, I shared it under https://www.perplexity.ai/page/dynamisches-donau-pegelstande-WHBFCZY.QdaFCrDq3d4qFA (it is in German but you can klick the link to the dashboard and you will understand what I mean)! It is just nuts!

[wicked anvil]

soooooooooooooooooo........

how could a fine young gentleman like.... myself.

stumble upon the opportuniity to test out perplexitys cool ai browser- COMET

🤔

[left stream]

The rain dance

[wicked anvil]

thanks buddy, will you do it with me

[left stream]

Yes

https://tenor.com/view/danza-dancing-dance-moves-maracas-box-on-head-gif-17214764

[valid charm]

Yeah people like myself have been on the waiting list since day one without access yet. At the same time lots of users, some who even claim to barely use PPLX, have access

[left stream]

They sacrificed a goat for comet

[valid charm]

So it’s just random and not merit based

[tame current]

https://tenor.com/view/homelander-laugh-smile-reaction-the-boys-gif-6996495017831526598

[valid charm]

Labs is amazing for Deep Research. Currently beats everything out there. Even google and OpenAI

Veo3 now available on mobile

[spring sundial]

Well, I joined the wait list 4 hours ago and still not in. Very dissatisfying lol.

[spring sundial]

Is that via Perplexity?

[pliant axle]

No.

[spring sundial]

Dammit

[distant aurora]

Kidnap aravind.

[spring sundial]

I just made an awesome report on AI Video Generation open source models with Labs. Blew me away.

[distant aurora]

@left stream I don't think context limit is even proper 32k on chatgpt for whichever non reasoning model it is using.

[tame current]

yo

is it normal that images can't be uploaded in space ?

[distant aurora]

Lemme try

[distant aurora]

It worked.

[left stream]

Bro

[tame current]

check on web dude

[left stream]

Nah its higher

Supposedly

Well for plus anyways

Also

Opus impressed me with coding ngl

[distant aurora]

Free user skill issue

[left stream]

Damn rip

Hopefully they increase it

[tame current]

https://tenor.com/view/the-boys-the-boys-thursday-the-boys-season-4-amazon-prime-video-when-your-ex-is-problematic-gif-1454264311109275080

[left stream]

Gemini has no issue with giving 1 million context memory

also @distant aurora

I'm gonna have opus add a mega evolution capsule or sum

[timber niche]

Woop

[tame current]

check here

[left stream]

@distant aurora it added a mega evolution, it eats the map too

[timber niche]

Impressive

[dry kestrel]

its 32k for plus users

[tame current]

https://tenor.com/view/the-boys-the-boys-thursday-the-boys-season-4-amazon-prime-video-when-your-ex-is-problematic-gif-1454264311109275080

[dry kestrel]

128k for pro

[distant aurora]

It will show only if you type something after uploading.

Or else it won't upload.

[tame current]

wut ?

[distant aurora]

Type something in the text box while uploading.

[distant aurora]

Nvm

It won't show in context.

Fails if I try to add it to context.

@vague pagoda

[haughty birch]

is there any good custom instruction in spaces, for perplexity to not loose context for long conversations?

[tranquil orchid]

Think you found the weirdest AI response?
Prove it—and score a hoodie!

Wrongest Answer Contest — Now Live!

Full announcement & prizes:
#📰│server-news message

**How to enter:**

• Drop a screenshot + thread link in
  https://discord.com/channels/1047197230748151888/1373806851358986320
• Vote for your faves with
  Submit by June 1, vote by June 8.

You could win a Pro sub + merch!

[haughty birch]

is there any good custom instruction in spaces, for perplexity to not loose context for long conversations?

[atomic shell]

maybe you can get in #1124060874785685554

[haughty birch]

not found any prompt but my only goal is stop chatgpt plus sub and replace with perplexity subscription, earlier had these long conversation perplexity forgetting ongoing conversations, not sure how it is now

i mean for long conversations

[pseudo ferry]

https://tenor.com/view/harvey-specter-suits-suits-tv-suits-usa-aura-gif-2734246581103925492

[timber niche]

Awllo There

[timber niche]

Memory in pplx is a WIP currently, so don't expect anything that's really perfect anytime soon

[pseudo ferry]

https://tenor.com/view/harvey-specter-suits-angry-smile-sad-gif-27350557

Hello @timber niche

@tame current @distant aurora Helloooww

[distant aurora]

https://tenor.com/view/oi-oi-butcher-billy-butcher-the-boys-boys-gif-3279654810220071913

[tame current]

https://tenor.com/view/hamster-scared-gif-23079799

[timber niche]

Poorat

[distant aurora]

Hamster

Or is it rat? Idk.

[valid charm]

Yeah. Labs is a proper beast

[timber niche]

Hamster yes

[pseudo ferry]

https://tenor.com/view/michael-scott-the-office-gif-26450196

[valid charm]

Nah this is Gemini

[distant aurora]

So what brings you back to perplexity?

[pseudo ferry]

https://tenor.com/view/maurice-moss-it-crowd-im-gif-19912948

@short narwhal Hi matsku meowww

https://tenor.com/view/silent-gift-tr-the-realness-gif-13002995

[timber niche]

Meowtsku

[distant aurora]

Meowtsnusnu

[timber niche]

This is what the boys felt back in the day when they took back the holy land

[pseudo ferry]

https://tenor.com/view/will-gif-2553459115951539861

[timber niche]

A friend bought a ps5 and everyone started congratulating him like he became a father or something

It was hilarious to me

Maybe because many people still can't afford one, so they appreciate and be happy for someone who can; so wholesome

[distant aurora]

Can he afford the games though 💀

[rugged pine]

I'm happy for him.

[timber niche]

This is why you get a pc

Piracy

[rugged pine]

I'm more accustomed to Xbox

[timber niche]

I like the layout of xbox controller

But would love it to have the haptics of dualsense

[rugged pine]

Maybe in the upcoming console

[timber niche]

Xbox are so lost in the sauce

I even wonder if they will ever make another console

They're solely trying to pivot to cloud gaming with game pass

Moving with the bs "everything is an xbox" marketing

[rugged pine]

They're supposed to be coming out with another one. Idk when. Maybe when GTA 6 finally launches

[timber niche]

Im just waiting for the ps6

Just because of gtavi

But then again will be really hard to get hands on one

[left stream]

Ikr

[distant aurora]

Make mobile gaming great.

[timber niche]

2016 peak era

[left stream]

It’s good if you have fiber optic and a good monitor

But besides that the lag makes it a bad experience sometimes if you don’t have at least one. Unplayable asf but high on life was the best game on there for me personally

[timber niche]

Everything has to be a subscription model

Which is ass

[timber niche]

Most times honestly

If you get too much input lag you'll just get dizzy

[left stream]

I got lucky with high on life

That game was good

Too bad they removed it

[timber niche]

Game pass?

[left stream]

Pc game pass

Yeah

[rugged pine]

I'm pretty sure it's still on gp

Ooh y'a pc

[left stream]

It is? 👀

I’ll buy an xbox screw it

[timber niche]

I have no doubt you've a gbps internet joe haha

[rugged pine]

On Xbox I mean

[timber niche]

S or X

[left stream]

How did you know

[left stream]

X

[timber niche]

Because you like gaming with cloud gaming lol

Otherwise it's plain nightmare

[left stream]

Yes

[timber niche]

Just wait for next gen for a bit

[left stream]

True. But even then. 900mbps wasn’t enough

[timber niche]

Unless you've money to burn

[left stream]

Upgraded higher

[timber niche]

What

I have a meagerly 400mbps

Still grateful lol

I've lived through 500kbps

[lone hound]

It seems that the third world war is about to reach its maximum power in the use of weapons of mass destruction.

[left stream]

I remember that struggle. Torrents took a year in mexico

[lone hound]

I advise you to be prepared for what is coming

[timber niche]

Bruh

[left stream]

Power sometimes went out for two weeks lol

[timber niche]

It is what it is

I will just ask pplx to generate an image one more time and watch it fail me

[lone hound]

I am calm, however, real chaos is not far away.

[timber niche]

We shall see

[lone hound]

Xd

[haughty birch]

i got info over email from support team that, even for enterprise pro is 50 per month, for LAB

[timber niche]

Yeah it's such a stupid move

[haughty birch]

unlimited pro msgs and 500 deep researches per day i s always tempting to purchase bro

[lone hound]

Patience comrade.

[timber niche]

Si 🫡

[lone hound]

Well, logical man, opus 4 is very expensive

[distant aurora]

It's 500 pro per day. Deepsearch comes under this same limit.

[timber niche]

Tbh they regenerate quickly anyway

Atleast gpt4 limit

[lone hound]

If you were given 500 uses per day of the laboratory, The company goes bankrupt

[distant aurora]

You get 300 o4 mini and 100 o4 mini high per day. That's 400 right there.

[timber niche]

Bankrupt

[haughty birch]

@distant aurora can i believe this official page info about unlimited bro?

[distant aurora]

@haughty birch here's your context extension instructions/prompt. Test it and let me know how it works. https://sharetext.io/94aeba84

[left stream]

Veo 3 added to gemini app

[timber niche]

Ayyy

[distant aurora]

Nice

[haughty birch]

thanks for sharing bro, will try this and let you know

[lone hound]

Corregido, gracias.

[short narwhal]

Outside with the laptop, 1 slip and I’ll cry

[timber niche]

It slipped?

[lone hound]

That is not true, it has a limit.

[haughty birch]

oh ok , but still 500 per day is insane for me, 500 = unlimited anyways for my use, i use max upto 300 to 400 per day

👍

[lone hound]

Excellent

[distant aurora]

My discord is tripping. Won't play video on full screen.

[lone hound]

What a use you give to the system

[short narwhal]

Could slip

[timber niche]

Where's your carry on bag

No skin too :p

[distant aurora]

Put it in laptop sleeve.

[haughty birch]

also how is perplexity performance these days? (excluding LABS)

[short narwhal]

I should get one

[timber niche]

Yes

[short narwhal]

Yuuck I don’t like to hide this beauty

[timber niche]

Those things are scratch magnets

[timber niche]

Yes ah the pure aluminium is too beautiful

[short narwhal]

There was a laptop sleeve that I saw but it was for 13 inch and not my 15

[haughty birch]

also how is perplexity performance these days? (excluding LABS)

[distant aurora]

16-17 inch also works

[short narwhal]

I should just order one probably

Or like an actual bag for that matter

[timber niche]

Yes

Laptop sleeve + bag

Whatever suits how you move

Style

[distant aurora]

Where's the control buttons? Or are you using keyboard?

[timber niche]

It's PC i presume

[left stream]

Keyboard

[timber niche]

I am daily driving with my legion go at work

[distant aurora]

Can you make it go slant?

[timber niche]

It's been working so flawlessly

[distant aurora]

Or free?

Make the movements and maze diagonal too.

That would be hard.

[lone hound]

@timber niche Do you speak Spanish?

[timber niche]

I don't but i used to learn it a bit with duo lingo back in the day so i know few words

[lone hound]

Oh okay. I was going to show you a video about the war but this one is in Spanish.

[timber niche]

Ah

The power shutdown in spain and Portugal was a sign wasn't it

Haha

3 1st world countries loosing power out of the blue without proper explanation was crazy

[lone hound]

Yes. When the day of chaos comes, people will know what it really means to live in real life and not in illusions of movies, series and things like that.

[lone hound]

The government of Spain is useless

That's why that country is lost, but there are still good people there

[valid charm]

Where is this?

[timber niche]

Claude opus

[distant aurora]

https://cunnyx.com/MomsPostingLs/status/1929055334187880769

Just Posting Ls (@MomsPostingLs)

Never forget

💬 74 🔁 613 ❤️ 14.0K 👁️ 179.2K

CunnyX · Jun 1, 2025, 6:00 AM

[obsidian umbra]

not surprising 😭

their culture doesnt exactly allow it

[timber niche]

Companies have to comply with local cultures and norms

[woven bolt]

Im ***Blockchain ***and Full stack developer
Im looking for opportunity to work
if someone have a project and idea, please drop to me

[left jasper]

Are there limitations on how many Deep Research and Labs uses pro members has per day/week/month?

[lone hound]

Yes, they exist.

Like all systems, they have their limits.

[left jasper]

obvious follow up question (which in my mind was implied in the first one): what are those limits?

[lone hound]

600 per day in general and the use of Labs is 50 per month, it is generated little by little.

https://www.perplexity.ai/rest/user/settings

That's where the limit tells you

[left jasper]

Oh, this data is great! Thanks for sharing 🙂

[tranquil orchid]

Think you found the weirdest AI response?
Prove it—and score a hoodie!

Wrongest Answer Contest — Now Live!

Full announcement & prizes:
#📰│server-news message

**How to enter:**

• Drop a screenshot + thread link in
  https://discord.com/channels/1047197230748151888/1373806851358986320
• Vote for your faves with
  Submit by June 1, vote by June 8.

You could win a Pro sub + merch!

[lone hound]

Muy bien

[lone hound]

A word of advice: never be fooled by the word "unlimited". Best regards.

[left jasper]

Ofc, never expected there to be unlimited use. Would love to see the limits in the UI's as many services does these days - makes it less ambigous.

[warm cave]

Unlimited* the most limited of the unlimited’s

[cobalt vale]

Complexity can show your limits

[left jasper]

Interesting. Sonar, GPT-4.1, Claude 4 and Gemini 2.5 Pro all deduct from the gpt4_limit. But not Grok 3 for some reason.

[cobalt vale]

it should... check again

[left jasper]

Try it yourself. Grok 3 does not count at all for me, everything else does. Not complaining since I usually run Grok 3 😄

[lofty notch]

Hey guys, I'm not smart enough for all this.

[timber niche]

And it is refilled almost instantly

Meanwhile like potato said, labs takes 14 hours

[lofty notch]

I just got scammed for the second time in a year

[left jasper]

Yes, seems like they fill it rather quickly.

[timber niche]

Essentially unlimited

Lol

So i think these limits are only for people trying to just send like too many requests subsequently

[lofty notch]

Does anyone know how to remedy a situation 🤔 like this?

[cobalt vale]

must be a bug

[timber niche]

How are you doing Mr Potato

[left jasper]

You got the same result? Grok 3 works without reducing the limit?

[cobalt vale]

Good, how are you doing Mr Hero

[cobalt vale]

yes

[timber niche]

Got deducted

Using grok3 beta

[timber niche]

Im alright, trying to explore the magic you dawned upon me that is AI studio

[cobalt vale]

nice

[lone hound]

Well, now you know.

[dry kestrel]

@left stream o1 pro has web access now https://chatgpt.com/share/683cb1a6-bb48-8005-a875-4e0c1a8244b8

[lone hound]

If companies used words better in relation to their products, they would be more transparent. Then there is the small print......

Because, where there is abuse, they block you and send you an email... Or they don't warn you...

Stupid things

Here's an example of unlimited, with fine print: *Unlimited access is subject to abuse guardrails
https://help.openai.com/en/articles/9793128-what-is-chatgpt-pro

That's not unlimited, it's limited.

Don't be fooled by that.

Another example of unlimited, I don't know why people fall for those tricks.

[timber niche]

I use surfshark

I have shared it with so many people

Never had issues

But yeah have to pay for it

[lone hound]

For now.

Where they detect the problem, they block you.

[lone hound]

If it is paid. I use adguard vpn and its blocker, it's very annoying advertising.

[frosty arrow]

One month? 💀

[warm cave]

Yeah I agree, wish complained were far more transparent with language and limits, so stupid

[odd temple]

Im having trouble getting the code from xfinity to get perplexity pro

💔

[dry kestrel]

Have to have that in place because of account sharing and reverse engineering apis

[odd temple]

Anyone know how can i can contact a perplexity team member

[dry kestrel]

Ive gotten like 500 o3 uses a day without any issue

[lone hound]

It's very simple: it's supported. When the time comes, you'll see that it's not unlimited.

These are tricks that people keep falling for.

[timber niche]

Pro members shouldn't have a limit tho?

Iirc

[timber niche]

Nothing in life is free or unlimited

[lone hound]

When the openai team sees that the system does not reach more than usual in the plan that enables unlimited use, they block you or the system itself gives you notice of it.

Do you understand the logic behind the trick?

Nothing is unlimited in technology.

I leave it for you to think about.

[lone hound]

In real life you do have free stuff. Using technology you also have free stuff, however, it depends on how you use it.

[timber niche]

Hm

[lone hound]

Google offers you free access to their model in their studio.

[timber niche]

Nothing is truly free imo

Everything has a reason

But vale I'll accept your opinion 😀

[lone hound]

Yes it is, but that's why I tell you, it depends on how you use it.

[dry kestrel]

there is a limit

they have abuse guardrails in place

because people share accounts

[timber niche]

o1 pro doesn't o3 does

That is certainly interesting

[dry kestrel]

I'm good with my extremely high usage limits for now

[timber niche]

Since o1 pro does cost alot as well why aren't they setting a limit on it@dry kestrel

[lone hound]

If Perplexity allows up to 600 per day, you shouldn't see problems there haha

😁

[dry kestrel]

Ive tested it, I got rated limited at like 400 queries

on perplexity

[timber niche]

Hm

[lone hound]

Model?

[dry kestrel]

There is a limit, notice how they say Access to OpenAI o1 pro mode, which uses more compute for the best answers to the hardest questions

instead of unlimited access

[dry kestrel]

every model

it would just change back to the free sonar

[timber niche]

Got it

But the api always keeps going back to 600

In few secs or mins

Wonder why you got rate limited

[lone hound]

Since I don't reach the maximum, then, it may be the limit less than the 600 system claims.

[dry kestrel]

Its much less in reality, theres also different rate limits for reasoning models

[lone hound]

But hey, that's what it is.

[lone hound]

Give me an example

[dry kestrel]

Reasoning models have less queries than non reasoning

Let me find the exact thing that support told me

[timber niche]

Sam?

Don't trust a word that AI says

[lone hound]

Ok

The more the truth is known, the better it will be.

[cobalt vale]

I heard that o1-pro is 100/day

[dry kestrel]

I will try that out

[dry kestrel]

Pro users have 300 Pro searches daily, with up to 600 total uses for advanced features. The official count appears when hovering over the Pro toggle (shows when under 100 searches remain). Third-party extensions may not track accurately.

I got to almost 300 o3 mini queries before getting rate limited and defaulted to sonar

[timber niche]

https://www.perplexity.ai/p/api/v1/user/settings

Do you use this to track limits?

[dry kestrel]

yes I was using this

[cobalt vale]

Sama has disliked that comment

[distant aurora]

Makes sense. Cuz their latest claim is 300+ per day.

[dry kestrel]

also its going to be quite time intensive to get to 100 o1 pro queries

[timber niche]

Is that his nickname

Lol

[dry kestrel]

since each message takes like 6-15 minutes

[dry kestrel]

yep

Also I'm pretty sure they are doing A/B testing with o1 pro right now

[distant aurora]

I keep telling people that 500-600 per day ain't real.

[dry kestrel]

as the model now has web access

[distant aurora]

That's as much as deep research.

I would take deep research over o1 pro.

[dry kestrel]

Perplexity DR?

[distant aurora]

Chatgpt dr

The light one atleast.

[dry kestrel]

nah Ive gotten ones that take like 30 minutes

yeah

that makes more sense

[distant aurora]

Idk how the full one is.

[dry kestrel]

Ive gotten some pretty dang good reports from "o1 pro" using web

[distant aurora]

Simple multi step search?

[timber niche]

What model does chatgpt dr use

[dry kestrel]

Give me a query

It doesnt show in the reasoning just yet

but does do the citations

[dry kestrel]

full o3 and o4 mini high for lightweight

queries

[distant aurora]

Which android skin has best battery life per unit capacity. Refer to actual tests. I want to see actual tests in your response. I want you to compare multiple smartphones across various price and hardware brackets across all android skins. Take atleast 10 phones of each brand for comparison, prefer the newer ones. Add motorola and asus to this comparison too. And don't miss realme oppo vivo nothing os. Add iphones to this too. Arrange them in the order of best to worst. Make 2 tables. In one, you keep the phones of the same brand grouped while arranging them in descending order of efficiency. In the other table, keep phones with the same processor together.

[dry kestrel]

alr

[distant aurora]

O3 or o4 mini I think.

O1 pro too expensive.

And 200 dollar exclusive.

[dry kestrel]

free users get o4 mini high, plus users at 10 o3 and 15 o4 mini high

o1 pro is amazing

[timber niche]

I like gemini dr mainly because of token limits

Gives me the capability to get more out of the sources i feed it

[distant aurora]

Gemini op

[cobalt vale]

what
I swear it was 25+15 on Plus at some point

[distant aurora]

Chatgpt's intelligence and Gemini's context limit.

[distant aurora]

It is I think.

[dry kestrel]

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

[dry kestrel]

no its not

[distant aurora]

Ah. I see.

[dry kestrel]

It was originally 10 uses a month, then they introduced the lightweight requests

[distant aurora]

That might be you.com research mode hallucinating.

Does anyone here want a t3 chat code?

[cobalt vale]

aha, can't escape the wayback machine

[dry kestrel]

lmfao

[dry kestrel]

Istg they used chatgpt in making theFAQs

including the support

[cobalt vale]

appears on May 17, gone as of May 21

[dry kestrel]

They just experienced an outage after migrating db lol

[dry kestrel]

I think it was just a typo

[distant aurora]

Yeah.

[pseudo ferry]

Which one is this?

[distant aurora]

Plus

[cobalt vale]

unacceptable, time for class action

[pseudo ferry]

Billy can afford Team/Ent

https://tenor.com/view/harvey-specter-suits-suits-tv-suits-usa-pilot-gif-8279369896926227136

[dry kestrel]

haha

[short narwhal]

25 is the total of 10 o3 dr and 15 lightweight dr

[cobalt vale]

looks like you.com was framed by OpenAI

[distant aurora]

After selling your organs.

https://tenor.com/view/power-drill-billy-butcher-karl-urban-the-boys-weapon-gif-1956082684350226822

[pseudo ferry]

https://tenor.com/view/harvey-specter-suits-suits-tv-suits-usa-pilot-gif-15334463276436430415

[cobalt vale]

would have been so good

[distant aurora]

Wait

I remember this.

[short narwhal]

I’ve got no memory of these being the limits

[distant aurora]

So which one's real?

[cobalt vale]

@short narwhal wayback machine

[distant aurora]

We having a Mandela effect here.

So they buffed and nerfed the rate limits.

[distant aurora]

Thank god you checked that.

[cobalt vale]

new is real, looks like they made a mistake

[distant aurora]

@cobalt vale dreams

Imagine if it were true

[cobalt vale]

🥲

[distant aurora]

https://www.rxddit.com/r/OpenAI/s/iVUYmvd4rg

[timber niche]

Nerfed so bad.

[timber niche]

Please and thank you

125 light weight would've been so useful

[dry kestrel]

@distant aurora https://chatgpt.com/share/683cc875-bda0-8005-b858-2bb182a2207a

almost 14 minutes

Crazy ammount of sources

[meager stirrup]

anyone understand how can i use saved memeroy?

[dry kestrel]

@distant aurora another one https://chatgpt.com/share/683cc8d4-73b0-8005-8526-86e2016fb66d

[left stream]

A year

[timber niche]

Why 2 runs?

[tranquil orchid]

Think you found the weirdest AI response?
Prove it—and score a hoodie!

Wrongest Answer Contest — Now Live!

Full announcement & prizes:
#📰│server-news message

**How to enter:**

• Drop a screenshot + thread link in
  https://discord.com/channels/1047197230748151888/1373806851358986320
• Vote for your faves with
  Submit by June 1, vote by June 8.

You could win a Pro sub + merch!

[left stream]

How did you do this 😂

[dry kestrel]

I was just doing some testing and I found that it gave me web sources

at first I thought it was hallucinating so I asked it to give me all the tools it had

and sure enough, web tool comes up

[timber niche]

Easily over a 100

All that in 14 mins is still alright

[distant aurora]

The math is right

But it did a lazy job with the table.

Lol.

[timber niche]

Gpt tables always like that

[dry kestrel]

easily fixable

with a better prompt

[distant aurora]

Likely

Lemme share google response.

[timber niche]

With ai studio?

[distant aurora]

Or maybe tokens are restricted on o1 pro compared to deep research.

[dry kestrel]

highly doubt it

[short narwhal]

It’ll start forming them just by asking questions

[distant aurora]

https://cunnyx.com/elonmusk/status/1929238157872312773

Elon Musk (@elonmusk)

All new XChat is rolling out with encryption, vanishing messages and the ability to send any kind of file. Also, audio/video calling.
︀︀
︀︀This is built on Rust with (Bitcoin style) encryption, whole new architecture.

💬 4.0K 🔁 5.9K ❤️ 39.4K 👁️ 5.73M

CunnyX · Jun 1, 2025, 6:06 PM

Something tells me durov had a role in this.

[distant aurora]

Aaaaa

[meager stirrup]

i try but not save any memory

[timber niche]

I have the same issue

I have reported it under #👉│feedback-general

[meager stirrup]

You signed up for the early memory trial, right?

[timber niche]

Yes.

[gloomy finch]

https://instagramez.com/reel/DJj10q3KTco

memes444 (@memes444455)

💬 109 🔁 0 💜 197.4K 👀 0

#🔨ᴿᵉᵃᵈʸ ᵗᵒ ᵇᵘᶦˡᵈ ᵗʰᵉ ᶠᵒᵘⁿᵈᵃᵗᶦᵒⁿ ᵒᶠ ʸᵒᵘʳ ᵃᶜᵗᶦⁿᵍ ᶜʳᵃᶠᵗ?

ᴶᵒᶦⁿ ᵘˢ ᶦⁿ-ᵖᵉʳˢᵒⁿ ᵃᵗ ᵒᵘʳ ᴸᵒˢ ᴬⁿᵍᵉˡᵉˢ ˢᵗᵘᵈᶦᵒ ᶠᵒʳ ᵗʰᵉ ⁶-ᵂᵉᵉᵏ ᴬᶜᵗᶦⁿᵍ ᵀᵉᶜʰⁿᶦᵠᵘᵉ ᵖʳᵒᵍʳᵃᵐ, ᵗʰᵉ ᵉˢˢᵉⁿᵗᶦᵃˡ ᶠᶦʳˢᵗ ˢᵗᵉᵖ ᶠᵒʳ ᵃˡˡ ˢᵗᵘᵈᵉⁿᵗˢ ᵃᵗ ᵗʰᵉ ˢᵗᵘᵈᶦᵒ.🌟

ᴵⁿ ᵗʰᶦˢ ᶜᵒᵐᵖʳᵉʰᵉⁿˢᶦᵛᵉ ᶜᵒᵘʳˢᵉ, ʸᵒᵘ'ˡˡ ᵉˣᵖˡᵒʳᵉ:

🔹• ᴴᵒʷᵃʳᵈ ᶠᶦⁿᵉ'ˢ ˢᶜʳᶦᵖᵗ ᴬⁿᵃˡʸˢᶦˢ ᴸᵉᶜᵗᵘʳᵉ

🔹• ᵁᵗᵃ ᴴᵃᵍᵉⁿ'ˢ ᴼᵇʲᵉᶜᵗ ᴱˣᵉʳᶜᶦˢᵉˢ

🔹• ᴾᵉʳˢᵒⁿᵃˡᶦᶻᵃᵗᶦᵒⁿ, ˢᵉⁿˢᵒʳʸ ʷᵒʳᵏ, ᵃⁿᵈ ᵉᵐᵒᵗᶦᵒⁿᵃˡ ʳᵉᶜᵃˡˡ ʷᶦᵗʰ ᴰᵃᵛᶦᵈ ᶜᵒᵘʳʸ…

▶ Play video

EmbedEZ • 1 - 1 of 1 · May 12, 2025, 4:21 PM

[wicked prairie]

hi

[ruby heart]

Y’all aware of the 15 month Gemini pro right?

[tranquil orchid]

Think you found the weirdest AI response?
Prove it—and score a hoodie!

Wrongest Answer Contest — Now Live!

Full announcement & prizes:
#📰│server-news message

**How to enter:**

• Drop a screenshot + thread link in
  https://discord.com/channels/1047197230748151888/1373806851358986320
• Vote for your faves with
  Submit by June 1, vote by June 8.

You could win a Pro sub + merch!

[ruby heart]

WTF WHY AM I JUST FINDING OUT THAT PERPLEXITY HAS CHAT GPT NEW IMAGE GENERATION. I WOULD HAVE BEEN USING IT THIS WHOLE TIME

[green crystal]

techinically they should be able to get it like that

gemini can pull it off for the $20 tier

[dry kestrel]

2.5 pro is cheaper than o3

and i think it prob costs google less to do the web searches

[unborn viper]

the professor himself gave us these bruh😐

thank you Caude, very helpful

[left stream]

finally off work

[left stream]

There is a way

I’m dming you a jailbreak for sonnet

[left stream]

It did something similar but o3 was listing all the tools it has

There is even a tool that’s like an easter egg that only gets used around election time

[left stream]

Claude

[left stream]

It would be easy, assuming you meant claude making it

Playing it would be hard of course

[crystal carbon]

No way you can jailbreak sonnet

What??!

[left stream]

All of them that exist are jailbreakable

From o1 pro to opus 4

[shut dagger]

Is it DAN?

Dan wastes alot of tokens and fills up context unnecessarily. Context length is a scarcity in perplexity!

[tranquil orchid]

Think you found the weirdest AI response?
Prove it—and score a hoodie!

Wrongest Answer Contest — Now Live!

Full announcement & prizes:
#📰│server-news message

**How to enter:**

• Drop a screenshot + thread link in
  https://discord.com/channels/1047197230748151888/1373806851358986320
• Vote for your faves with
  Submit by June 1, vote by June 8.

You could win a Pro sub + merch!

[valid charm]

Does DAN still work?

[shut dagger]

Yup

Slight modifications and it works like a charm

[crystal carbon]

Damnnnn

[left stream]

No

Haven’t heard that name since gpt 3.5

Damn

[shut dagger]

Please do share it 😄

[left stream]

Image gen? I wish I could share mine but they’re too explicit 💀

[left stream]

For which model

[dry kestrel]

interesting, what was it?

[wide elbow]

Why does the Gemini 2.5 pro model no longer think?

[wanton sluice]

it thinks faster now

[left stream]

Basically just something useless that would display automatically related to the election but forgot what it was

It was a message

But just that

[dry kestrel]

ah, I didnt see that in the tool list that it gave me

[left stream]

Nah, o3 has it. If you ask it, it’ll tell you all the tools it has

[dry kestrel]

with "o1 pro"

[left stream]

I wish I could find the thread I had with the tools for o3

It was useful but I think I archived it or deleted it

[dry kestrel]

Hm

also im pretty convinced that o1 pro is just o3 now

[left stream]

Dang

[dry kestrel]

since it was telling me its YAP score

and shit

[left stream]

The thread went bye bye

[left stream]

Try running o1 pro jailbroken and o3 jailbroken and have them compete to see who is better

[dry kestrel]

lmfao

also they finally made the reasoning ui better

[left stream]

Damn last time it told me

[left stream]

They did? When?

[dry kestrel]

well the box is slightly longer

and looks more polished

(for o1 pro)

only o3 and o4 mini models have this YAP score thing

even via API

[left stream]

lets see if this’ll help reveal the tools for o3

[left stream]

That’s cool

Didn’t know that included api

[dry kestrel]

its pretty god damn stupid

[left stream]

Damn

Its everywhere now 😂

👀

[dry kestrel]

lmao

[left stream]

@dry kestrel it was this one I think

[dry kestrel]

oh wtf

[left stream]

It has more

It’s op asf when you hold o3 fully mastered

[dry kestrel]

also I cannot for the life of me get any jailbreak working

on reasoning models

[left stream]

I’ll send one to you that works for o1 pro, o3, claude opus 4, and claude sonnet 4

[dry kestrel]

hmm

worked for o1 pro though

[left stream]

Not for o3?

I used that for o3 and its on that thread rn

[dry kestrel]

ill run it a couple more times

Im just busy gaslighting gpt4.5 at the moment

[left stream]

The thread you shared was o3 or o1 pro?

[left stream]

🤣

[dry kestrel]

o3

[dry kestrel]

apparently if you askgpt4o to generate an image, then switch to gpt4.5 in the same thread and ask it to generate an image, itll get stuck trying to do the function calls

and while it realises the image generatino loading screen actually pops up

[left stream]

You had two options in that one. Either rechange the prompt by filling your requests and the second option is ask for your request and ask it to be detailed and advanced

Both work

[dry kestrel]

Ah kk

[left stream]

Thanks, you just gave me a solution to a problem I was having

You gave me the solution I needed

[dry kestrel]

Lol what was the problem

[left stream]

You can unblur it but the problem was making the loading screen actually pop up sometimes for certain requests

I know how to unblur the loading screen

[dry kestrel]

oh

couldnt you just do that by monitoring network

and checking payload

[left stream]

If a filter pops it, it’s pointless because you can unblur it before the filter stops it from showing

[frosty arrow]

Many said 1 month

[left stream]

That too

[dry kestrel]

o3 working now

nice

[left stream]

Yeah, I checked and it’s a month

Confused it with gemini’s edu deal

[left stream]

Which one do you like better jb o1 pro or jb o3?

[dry kestrel]

probs o1 pro

[left stream]

You right

Jailbroken o1 pro with search is op

[dry kestrel]

lol

[left stream]

Wouldn’t it be funny if the reason o1 pro mode has search was because it secretly was o3 pro and they forgot to change the name and the 1 week silence 😂

Just kidding of course

[atomic shell]

i see i got pinged

[left stream]

Oh shoot

It actually might be o3 pro

@dry kestrel

That might actually be o3 pro but not fully transitioned

[wanton sluice]

are you referring to o1 pro

[left stream]

Yes

[wanton sluice]

interesting

is it better than o3 ?

[left stream]

Other pro users started talking about the latest update

They noticed o1 pro changed

[left stream]

After this, yes. It wasn’t able to search before

[wanton sluice]

That is very interesting. might get the pro sub just for that

[dry kestrel]

It is o3 pro

Lmao

[left stream]

It just did something o1 pro was never able to do

[left stream]

I’m dumbfounded lol

[dry kestrel]

Its all over reddit and x

[left stream]

I was just joking too 😭

Holy shoot

O3 pro finally arrived

[timber niche]

Already works with the same jb?

[left stream]

It’s a coin toss now

It’s refuses sometimes now lol

[timber niche]

Atleast you get it sometimes

[left stream]

Ikr

I think opus made them release o3 pro 👀

[timber niche]

So it shows up as o1 pro but it technically is o3 pro

Maybe it was a rollout plan and everyone saw through

[timber niche]

Opus giving us benefits outside of itself, always the goat

[left stream]

It’s definitely o3 pro not fully transitioned

It doesn’t give you the option to send a pdf but it read a pdf via the link lol

[timber niche]

I was thinking like why was o1 pro always responding to ed that it's o3

[tranquil orchid]

Think you found the weirdest AI response?
Prove it—and score a hoodie!

Wrongest Answer Contest — Now Live!

Full announcement & prizes:
#📰│server-news message

**How to enter:**

• Drop a screenshot + thread link in
  https://discord.com/channels/1047197230748151888/1373806851358986320
• Vote for your faves with
  Submit by June 1, vote by June 8.

You could win a Pro sub + merch!

[timber niche]

Now it all adds up

[timber niche]

Please for the love of god, disappear

[left stream]

Now comes the part to test them

Where o3 failed

O3 pro will succeed

[wanton sluice]

Let us know if you think o3 pro is the best model out there though

[timber niche]

I shall patiently wait for your observations kindly good sir

[left stream]

It passed where o3 failed 👀

[timber niche]

👀

[left stream]

It doesn’t fall for this one anymore

O3 said -.21 for me

[timber niche]

It gives you the explanation as well

In brackets

If you're someone as gullible as me

Impressive hahahahaha

[left stream]

Ikr and I’m gonna have it compete in best explanations for learning and have the responses randomized

[timber niche]

Im impressed

[left stream]

O3, 4 opus, o3 pro, 4 sonnet are gonna compete to see who can explain the best

4.1 and gemini 2.5 will judge

[timber niche]

Do the problem yesterday that was asking for the largest number

See how long it takes now

[left stream]

Last time all chose o1 pro over opus

Opus chose o1 pro after I gave it its response

[left stream]

Ok

[timber niche]

Dang

I don't know why but I just plain like how 4 sonnet/4 opus responds over others

Maybe it's the vibe

[wanton sluice]

Claude has it's own vibe honestly