💬│general

[quaint ravine]

or gained actual money?:

from doing that?

[young hollow]

My Life has Certainly got Better Because of Ai

[quaint ravine]

i see

[young hollow]

Although im not a Expert Yet

[quaint ravine]

seems vague

[young hollow]

Im Pretty Certain when i Become A Expert i can Get Alot More Out of Ai

But im just not there yet

[quaint ravine]

yea i'm more curious how it actually improved actual measurable real world statistics for example

did u literally earn money that u can spend on a new car etc

[young hollow]

As far money is considered i haven't made any money with it because Im not a expert for people surely make money with ai

[quaint ravine]

or did it get you much higher marks in real time assessments etc

[ripe tulip]

I mean I can provide time savings just not specifically earned money but certainly money saved for my company.

[young hollow]

I predict i will surely earn money in future with Ai

[quaint ravine]

yep I would have the same experience

[ripe tulip]

One of my projects is to do with data dashboards for the national health service so depending how they use that data

[young hollow]

Time is Still Money Tho

[ripe tulip]

could argue impact on community care

but Im not that far into that project yet

[quaint ravine]

interesting

[young hollow]

Yup

Very Interesting

Ai can Write Content

So that's great

[quaint ravine]

i know what AI can do lol

I just wondering how it actually translates in real life for various people

[young hollow]

Someone here earns 20k+ $ A Month Writing Books and Novels

With Perplexity

[odd forge]

yo wt

[ripe tulip]

cool BUT WHY USING PERPLEXITY?

surely thats harder then just using a proper AI for it

[mental spade]

Hi

[odd forge]

what are the best ways to earn money using ai

[young hollow]

Is Very Subjective and Preferences Depends on Needs of One

[ripe tulip]

I was considering making a novel though I can;t even come up with the intial idea so I switched to domain flipping or learning about it atm

[young hollow]

Idk it just Works Well for Him

[young hollow]

I would write anime novels if i will

[quaint ravine]

k

[young hollow]

Tbh i would just write whatever can make most money

[odd forge]

?

[young hollow]

Now I Have Both Gemini and Perplexity Pro

What do i do LOL

[green crystal]

get openai and anthropic for the full collection

[young hollow]

Nah I can't pay for those

[quaint ravine]

Claude for my use case has been very disappointing

[young hollow]

Claude is Just for Coding and Writing...

[quaint ravine]

a lot of pubmed articles atually test Claude in Medicine etc

showing it performs high

tests and actual real-world real-life tho

aren't the same thing imo

[young hollow]

Yup

You can't Trust Anything Expect Your Practical Experience

[quaint ravine]

yea hence the paradox

I think anyone who has experienced at least moderately with AI

is starting to realise this

the issue is that many people seem to put universal statements when they haven't actually experienced all the use cases

[young hollow]

Exactly

[quaint ravine]

but to a certain degree, the AI will be making u do more work...especially with learning curve included etc

[young hollow]

That's A Paradox

Yeah Ai itself is Also A Skill

[quaint ravine]

yea

i think a low proportion of people will genuinely be able to adapt

and optimise its use

[dark quest]

i am wondering does perplexity api offer llama 4-scout or maverick model ?

[dim horizon]

Why does perplexity talk to me like its my secretary I want it to talk like its a human like chatgpt does
Where can i give it instructions on how to act

All i found was a place to describe myself not how i want it to talk

Or did i misunderstand the point of it for when i was subscribing

[distant aurora]

Make a space

Put custom instructions there.

Also, make sure to disable web search in your chat if you just want a chat buddy.

And pick the right model.

[dim horizon]

What do u recommend, I have chatgpt plus but its a family subscription so i dont get alot of freedom with it

So i took the perplexity paypal offer just to try it and so far i like the app itself more but my only problem was what i said

[young hollow]

Very Low Percentage Indeed

@languid star Can You Lend A Family Slot to My Brother?

[languid star]

gib email

[young hollow]

U won't ask for anything?

[young hollow]

Give me sometime let me ask my Brother Which Email he Uses Mainly

I swear if i knew you would had given me i wouldn't had paid 200₹

[dusky sun]

Hey Gen Z what's up

[vast shard]

sky

[dusky sun]

Let me ask something

[vast shard]

k

[languid star]

https://tenor.com/view/francis-boogie2988-flip-rage-gif-9160180

[dusky sun]

Where are you from

Nepal

#nepobaby

[abstract mesa]

2000 years later...

[lone raven]

hi

[pseudo ferry]

https://tenor.com/view/netflix-and-chill-cat-kitten-cute-relaxing-gif-16654308

[trail folio]

Job Application

@vague pagoda job application

[pseudo ferry]

https://tenor.com/view/harvey-specter-escathon-suits-angry-sad-gif-27350469

[trail folio]

Yoooo

[hard pewter]

Yoooo

[golden palm]

[dim horizon]

Anyone wanna slide comet

[modern sinew]

https://tenor.com/view/job-job-application-jobless-gif-2757097081210871087

Here we go chill guy

[trail folio]

[earnest latch]

hi

[strong shale]

@halcyon coral can i have pro role?

[halcyon coral]

Please check #💬│general message 🙂

[raven river]

[tall rain]

I’ve never tried labs what does it do?

Can I use to study?

[languid star]

https://tenor.com/view/breaking-bad-chemistry-experiment-scientific-walter-white-gif-17820266

https://tenor.com/view/cocaine-drugs-gif-23720718

[royal marten]

Is there Chat Memory for everything like chatGPT has and is it possible to set memories?

because if I ask it to memorise something, it doesn’t save it in memory

[tall rain]

Yea in settings

In the personalize tab

[golden palm]

wondering what the prompt was

[royal marten]

only delete no „add“ button

[tall rain]

Just type what you want to add and type “save this to memory” at the end

[hybrid grotto]

That's my homescreen, I use perplexity in general and gemini nano banana for images so much

[languid star]

Mine

[serene sand]

deep dive ➤ #💬│general message

🪼

[lapis token]

Hi, I would like to chat with a Perplexity Pro Support agent regarding billing issues.
Would anyone be able to help?

[languid star]

nope, no one is here regarding support

[lapis token]

Where can I get in touch with one?

[serene sand]

for billing contact [e-mail] or intercom pro support https://www.perplexity.ai/account/details

[lapis token]

Thanks Mars, nonetheless, my intercom support does not work.
I click, nut I does not do anything.

[serene sand]

which device, OS, browser

[slate flume]

yo how do I report users in this server?

there isn't a ticket channel

[serene sand]

you can DM @vague pagoda

[slate flume]

ah ok, ur not able to do anything?

[deft pulsar]

what's the image generation limit via chat gpt5 in perplexity pro?

[serene sand]

or report here..

[slate flume]

cant i just dm u?

it's a sensitive report

suspected pedophile activities

[serene sand]

please DM Kesku

[slate flume]

alright i hope he or she respond asap

i believe he is preying on multiple people

[twin grail]

....

yikes

[golden palm]

use 'pedo' to avoid timeout

[twin grail]

u can report here as the mod can straight up ban him

or u can dm @vague pagoda

[slate flume]

ye dmed kesku the evidence already

[golden palm]

give prompt

[languid star]

daym, who harassed u

[twin grail]

pedo ig

[slate flume]

he's going around dming people asking to become their friend

he clearly states that he's 33, ive told him that im 16 (im not underage btw) seeing what is his intentions and he told me to send a pic of myself

[lapis token]

I tried on:

• Desktop: Acer Intel(R)Core(TM) i7-4770K CPU @ 3.50 Ghz 3.50 Ghz.
• Laptop: HP EliteBook 840 G5

Browsers used both default and incognito modes (only the Bitwarden extension was active in incognito),

• Google Chrome : Version 140.0.7339.128 (Official Build) (64-bit)
• Brave Browser : Brave 1.82.166 (Official Build) (64-bit) Chromium: 140.0.7339.133
• Opera Browser : Opera One (version: 119.0.5497.141)

[young hollow]

@languid star Please Check DM

[viscid plover]

HI everybody

[languid star]

u too

[viscid plover]

Hi

[serene sand]

you have subscription?

[golden palm]

sounds gay

[slate flume]

oh i told him a girl too

[tranquil orchid]

5. Keep discussions relevant.

Stay on-topic and helpful in each channel. Use ⁠#💗│sharing to post tips, content, or cool discoveries.

[languid star]

https://tenor.com/view/how-are-you-doing-homelander-antony-starr-the-boys-whats-up-gif-26103434

[golden palm]

he is a girl ?

peak climax

[slate flume]

so ye he didn't backed off despite me telling him i'm 16, went to ask for my gender, and when i told him a girl, instantly asked for a pic

[slate flume]

no he asked for my gender, i told him im a girl

[golden palm]

you = femboy ?

eww

[slate flume]

no lmao, it's to see his response

[golden palm]

tomboi solos >

[golden palm]

ow

[slate flume]

anyways he should've backed off the moment i told him im 16 (im not 16 btw)

[golden palm]

got it

[golden palm]

aren't 16 yo kids humans too ?

[slate flume]

he stated that he is 33

a 33yo should NOT be talking to a 16yo especially on discord

[golden palm]

maybe looking for a young tech mind for his company

[lapis token]

Yes, I do.

[neat heart]

hi

[golden palm]

you missed internship bro

[slate flume]

he's clearly going around dming random people in this server and trying to find minors

[serene sand]

thank you for reporting

[golden palm]

is he @quick kiln ?

[slate flume]

yep

did he dmed u too?

[golden palm]

ya

[neat heart]

hi

[slate flume]

mods need to take actions fast

because some dumb kid

might just send a pic to him

[golden palm]

third hi soon cause no reply (this one will not count)

[lapis token]

Sure thing!

[neat heart]

hi

[golden palm]

mods are kids too btw

[modern sinew]

Ayoo

[slate flume]

wha-

bruh

[modern sinew]

They aren't

[golden palm]

hola

[golden palm]

~ msg was based on way of talking

[modern charm]

hello everyone! can please anyone tell how to return stepts in ai answers?

like that

i do not want search, i need only steps, but i don't know how to return it

[modern sinew]

You only want to see the steps right?

Click on this button

To get em

[languid star]

https://tenor.com/view/indeed-i-love-you-concur-thats-what-she-said-is-how-speed-gif-24958898

[golden palm]

guys is this a bug or network error ?

[modern charm]

yeah! but in this space there is no section with steps

a little bit frustrating

[languid star]

bug

[modern charm]

oh okay

need to make new thread...

[golden palm]

system bug or temporary ?

[languid star]

browser bug

[modern charm]

temporary other threads are okay idkw

[golden palm]

time to switch to chrome back (until this one get solved)

[golden palm]

huh

[modern charm]

okay thanks everyone, i am glad it is just a bug

[modern sinew]

Yes you can click on the same button to get the steps

They only come when steps are produced

As some responses can only have searches

[untold wave]

I got access to comet, it’s amazing. Some of you guys have feedback as well?

[modern sinew]

its annoying for me

Will rather stick to basic perplexity, i am too used to Perplexity + google wont be converting anytime soon

[languid star]

i left comet due to lack of vertical tabs 😌

[modern charm]

it is strange, but now anytime i am trying to make a new thead there is no steps always image in the third section...

[wary bone]

I found out that perplexity pro is not worth it
ChatGPT is better
It is not generating accurate output sometimes
What are your opinions?

[golden palm]

skill issues

[lapis token]

Each has its strengths and weaknesses.
Balance both and be unstoppable.

Add in Elicit.com and become even greater.

[languid star]

look at this lol

[golden palm]

imagine saying monitor is not good cause mouse's dpi is low

[golden palm]

mac or linux ?

[languid star]

windows

[modern sinew]

Thats a freaking windows

[slate flume]

wait how

[modern sinew]

Hammy has Apple version of discord

[lapis token]

That was a good one, haha.

[wary bone]

But it's providing big answers to MCQs and
I generated some codes which were not working properly

[slate flume]

ye i think i will grab chatgpt plus

[modern sinew]

+1 chatgpt is good for studying

[golden palm]

ow

[lapis token]

As long as you can afford...

[languid star]

https://tenor.com/view/antony-starr-the-boys-the-homelander-i-can-do-whatever-i-want-gif-17749198

[slate flume]

is that a custom client?

[languid star]

its a browser

[gloomy cipher]

Is this modded or nitro?

[slate flume]

ohh

didnt notice ur using a browser

[lapis token]

Remember, Perplexity offers a 1-year pro plan on Samsung devices and provides a month free for students, followed by $4.99/month as a student instead of the regular $20/month.

[modern sinew]

Thats cool which browser is it?

[green crystal]

what the

[golden palm]

name = ?

[languid star]

ZEN BROWSER WITH NEBULA THEME

[slate flume]

but its so much worse compared to chatgpt tho

but sadly chatgpt dont have student plans

[wary bone]

What if i don't use Samsung?

[golden palm]

perplexity is only for on live time researches and gpt is a chatbot

[lapis token]

https://tenor.com/view/idk-sassy-toddler-i-dont-know-cute-gif-16934315

[wary bone]

Kinda true
But it is generated biased answers to questions

[golden palm]

connect paypal account and get 1 year

[lapis token]

But seriously, grab a Samsung, and that's it.

Here is the full process.
https://www.perplexity.ai/search/samsung-galaxy-perplexity-pro-zQA4._ydRU2aI.Hw.epghA

[gloomy cipher]

Homelander

[golden palm]

non tech person detected successfully ->
opinion rejected brutally

[languid star]

https://tenor.com/view/homelander-angry-antony-starr-the-boys-nafkelah-gif-25554741

[wary bone]

I already have the 1-year subscription but i think it's not worth it

I got it for Coding purposes
But it's generated codes that don't work sometimes

[lapis token]

“Much worse” is a bit of a stretch.
Regarding research, it's a pioneer at the top of the LLM industry.

[gloomy cipher]

Thats new

[lapis token]

Did you use "Claude" agent?

[wary bone]

Yeah.
But in big codes it's still doing mistakes

ChatGPT also does many mistakes
But the number is lower

[still turtle]

[lapis token]

I see.
Let's hope that both agents make fewer mistakes over time.

So, does anyone know how I can contact support on Discord? Or are we bound to email if the intercom is down?

[lapis token]

Please, which browser is this?

[slate flume]

so is chatgpt plus worth it?

[lapis token]

Yup, it definitely is.
Fine print: depending on your use case.

[slate flume]

mainly just to understand math concepts etc

[spark pike]

Hi guys, I am new, ow can i know what AI to use on my case scenario?

[distant aurora]

https://vxtwitter.com/legit_api/status/1966483848302112806

ʟᴇɢɪᴛ (@legit_api)

↪️ Replying to @legit_api

originally it would consistently make beautiful and high fidelity simulations of a rotating earth-like planet (0 shot)

but then a few days later it would consistently make sub-par and terrible attempts https://t.co/pgVOXLwuS9

vxTwitter / fixvx · Sep 12, 2025, 12:47 PM

@green crystal Gemini deepthink got nerfed 😭

[spark pike]

Like what's best for solving uni math problems or research?

is it fine to use the research pro?

[distant aurora]

Gpt 5 thinking/O3. Use research for more thorough answers.

[weary abyss]

gpt 5 thinking

damn good

[slate flume]

does the free plan get gpt 5 thinking or gpt5

[weary abyss]

pro

[slate flume]

ohh so only pro plan gets gpt 5 thinking

[weary abyss]

gpt 5 thinking pro only i think

[green crystal]

lmao google is supposed to have crazy compute resources right
why nerf it
a $250 model of all things

[slate flume]

im talking bout chatgpt plan btw, not perplexity

[lapis token]

On problems, ChatGPT and Perplexity have different approaches and, sometimes, different answers.
On research, Perplexity has more reference links available.

[slate flume]

coz perplexity one is pretty ass ngl

[weary abyss]

hell nah

tf

gpt 5 thinking in perplexity is solid

[slate flume]

like compared to the actual chatgpt one

[spark pike]

Okie thanks for the info

[weary abyss]

along with claude

[green crystal]

chatgpt

[slate flume]

the actual one can break down concepts into simple understandable terms

[lapis token]

Always!

[green crystal]

the models have all their tools available

[weary abyss]

tbh i think it's better (excluding context window)

[slate flume]

whereas perplexity one doesnt go into detail

[spark pike]

How about the Research and Labs option in perplexity?

[weary abyss]

labs nice, research mid

[weary abyss]

claude for details

[spark pike]

chatgpt or them?

sorry for the questions T_T

[weary abyss]

chatgpt

[spark pike]

im just hella overwhelmd

[slate flume]

like i asked the same question to both chatgpt (free) and perplexity (using gpt 5 thinking)

chatgpt was able to explain the whole concept using tables, images as visualisation, asked followup questions to test my understanding, whereas perplexity returned 5 sentences

[green crystal]

perplexity doesnt give the models their native tools

[foggy dock]

whaaaat

[slate flume]

so ye perplexity one is honestly pretty bad

[serene sand]

deep dive ➤ #💬│general message

🪼

[spark pike]

Thank you

[weary abyss]

i mean

you dont find 10+ ai models for free

[green crystal]

its good enough for search but the rest is meh in comparison

[weary abyss]

elaborate

[slate flume]

ye for search it's good

it provides citation, sources etc

[weary abyss]

i use it for math daily

[slate flume]

bro try out chatgpt

[celest edge]

quality is better than quantity

[slate flume]

i think u feel perplexity is better because u havent seen the quality of chatgpt

[weary abyss]

if you know how to use the models correctly

[slate flume]

like im not kidding u

ive asked the same EXACT question

[weary abyss]

chatgpt doesnt explain much

[slate flume]

idk ur trolling or what

ive tested them for 2 days

[weary abyss]

?

[celest edge]

dckrider among us

[slate flume]

asked them both the same exact questions

[weary abyss]

relax lil boy

[slate flume]

10/10 chatgpt always return detailed answers

[weary abyss]

chatgpt dickriders

[celest edge]

shut your bic ass up

[slate flume]

but to give perplexity credit, it got one of the question correct whereas chatgpt gave a wrong answer

[weary abyss]

10yo behaviour

[slate flume]

but in terms of detail and understanding concepts, especially on math topics

[weary abyss]

YOU CANT SAY PERPLEXITY like it has alot of modules

[slate flume]

chatgpt crushes it every single time

[weary abyss]

that sounds stupid

[celest edge]

This guy's use case is pleasing himself via AI

[weary abyss]

chatgpt is meh. i only like gpts

[slate flume]

why don't u just give it a test rn

ask a question on euclidean algorithm for example

[celest edge]

2.5 pro beats the shit on perplexity to ground

[weary abyss]

i use it daily

[slate flume]

give it a try now

[weary abyss]

finally i can agree with this kid

[slate flume]

ask this prompt

"explain the concept of euclidean algorithm"

[celest edge]

might as well suck my dck

[weary abyss]

okay, lets try a math question

[slate flume]

ask the prompt i gave

[weary abyss]

i dont swing that way sorry

you seem gay but i am not

[lethal plover]

comet is pretty cool lads

[weary abyss]

how abt i try a math question

[slate flume]

how about u just ask the prompt i gave

[weary abyss]

how about you stop acting like a lil btch

[slate flume]

damn how old is this dude

[celest edge]

fucking 3 i swear to god

[slate flume]

anyways its alr, i've asked it, here's the results

[celest edge]

ill piss on his face before he can realise

[tranquil orchid]

3. Treat community members with respect.

No harassment, hate speech, or malicious behavior. Keep things positive and constructive.

[green crystal]

use thinking

[slate flume]

im on free plan 💀

[weary abyss]

gave me a half assed reply

chatgpt doesnt rock

[slate flume]

chatgpt was able to give a simple & concise explanation, gave an example and asked if i wanted a followup question

[weary abyss]

is sam altman your wife or smth

like stop galzin

[slate flume]

perplexity on the other hand decided to start venturing into pseudocode, does not simplify things down, instead returns a long chunk of text

[slate flume]

do u have eyes, are u able to read?

[weary abyss]

how abt you show us

[slate flume]

are u blind?

[weary abyss]

what model lil boy

[tranquil orchid]

3. Treat community members with respect.

No harassment, hate speech, or malicious behavior. Keep things positive and constructive.

[slate flume]

gpt 5 thinking

perplexity got absolutely demolished in this example alone

went off topic

returned long chunk of text

[weary abyss]

lemme see

[slate flume]

at this point, i might as well go read a book about euclidean algorithm

[celest edge]

lol perplexity cant even get basic physics questions right

[slate flume]

and not to mention, took a few mins too

[weary abyss]

it got it correct for me

maybe u modified it so it appears like this

[slate flume]

show a screenshot?

[weary abyss]

of?

[slate flume]

u asking perplexity the prompt??

u said it got it right for u

so show us?

[weary abyss]

it got it detailed and right yes

[slate flume]

so show us?

[slate flume]

again, it went off topic

u asked for euclidean algorithm, it went on to explain extended euclidean algorithm

when u look at chatgpt, it simplified it down to ONLY euclidean algorithm, then asked if you wish to learn about extended euclidean algorithm

[weary abyss]

elaborate

you probably gave it some command beforehand

[slate flume]

and even the step by step example isn't as detailed as chatgpt?

[weary abyss]

you cant cheat

[slate flume]

??

are u seriously this delusional?

[green crystal]

ah gimme the prompt i'll try in a bit

[weary abyss]

you can inspect element and change content 😂😂

[slate flume]

"explain the concept of euclidean algorithm"

im pretty sure mine was on gpt 4, coz i ran out of the limit for gpt 5

and yet it still crushed perplexity

[green crystal]

all in all chatgpt is the main model provider
the full model spec with all chain of thought tools are used

[slate flume]

there's simply no comparison here really

[weary abyss]

yeah if you cheat

[slate flume]

does chatgpt plus get gpt5 thinking btw or it's limited to chatgpt pro

[weary abyss]

xdd

[weary abyss]

you dont even know xdd

like actually crazy

[slate flume]

o it does

ye i might actually just get chatgpt plus

[weary abyss]

top 10 glazers and dont understand what hes glazin

[slate flume]

so when i show actual evidence, all he can do is say im cheating

[weary abyss]

those photoshoped pics xdd

2/10 ragebait

[distant aurora]

This was og https://vxtwitter.com/legit_api/status/1966483841037521023

ʟᴇɢɪᴛ (@legit_api)

Procedural planet simulation

by Gemini 2.5 Deep Think IMO https://t.co/ZEahxxBIUY

▶ Play video

vxTwitter / fixvx · Sep 12, 2025, 12:47 PM

[languid star]

https://github.com/0xk1h0/ChatGPT_DAN

[slate flume]

how do u even use older models tho

[languid star]

use prompt in chatgpt whatsapp

[green crystal]

was this prompt tested with 5 pro and grok heavy?

[distant aurora]

I don't think so.

I got the pagoda one btw in case anyone wants to try.

[weary abyss]

grok heavy?

isnt that 300$?

[distant aurora]

https://www.vxinstagram.com/p/DOihVeQAQk4

[minor sequoia]

What is this for lmao "hey guards here is how we do jailbreak" haha

[green crystal]

https://sharetext.io/032ce664

[green crystal]

yeah

[slate flume]

ah the formatting got ruined tho

[weary abyss]

it must be really good

[green crystal]

ah i forgor to mention that what you're seeing is 5 pro lol
making the 5 thinking version now

[slate flume]

ohh the best model?

[nimble creek]

Anyone alive right now?

[green crystal]

https://sharetext.io/6a5ebfa0

this here is 5 thinking

[green crystal]

yep

[slate flume]

ic

ok looks like their best model includes pseudocode too like what perplexity did

interesting

[green crystal]

stonks moment

[green crystal]

interesting

[fast meteor]

radio is back ⁉️

[slate flume]

btw out of all ur AI subs, which one is the most worth it in ur opinion

[fast meteor]

This live stream recording is not available.

[green crystal]

chatgpt pro

its the only one i got lol

[serene sand]

try now

[slate flume]

oh i thought u got grok

[green crystal]

nah

pro is expensive though i managed to use it for business

[slate flume]

ah

ye pro is too overkill for me

[green crystal]

and i got my sub money back pretty quickly

[open charm]

hi guys how are you doing

[distant aurora]

@languid star @gloomy cipher @young hollow https://www.vxinstagram.com/reel/DOia6uFEux5/

[green crystal]

unless you use agent and 5 pro allot then plus is good

[serene sand]

Tune In, Turn On, Drop Out

[green crystal]

my needs barely scrape by for what 5 pro is actually better vs 5 thinking lol

[languid star]

https://tenor.com/view/cat-meow-silly-cat-cat-explode-yellow-cat-gif-12654556298327310612

[slate flume]

i see

[slate flume]

how huge of a difference is it

[distant aurora]

Imagine if perplexity feed had a scrolling section for ai generated memes.

[green crystal]

if its for accurate math, analysis, comparison of data and indepth planning then 5 pro is leaps beyond 5 thinking

in a single prompt that is

5 thinking can pull it off but you'll have to guide it more often

[slate flume]

ah damn

but ye too much for me lol

even plus is pretty expensive already

[timber niche]

When is pplx remodel for android coming

Or do they not care about us

Lol

[green crystal]

pplx liquid glass theme when?

[open charm]

always give my laptop dad after switching windows if he saw linux , they will kick me from home , and property papers

[regal dawn]

Anyone taking part in SIH

[hasty salmon]

hi

[silent ocean]

imagine if perplexity shipped their redesigned app to android.

[topaz spindle]

Looks like this group does not have any perplexity support folks

[vernal brook]

Hello guys

[thin lake]

@regal dawn that’s it. i’m doing being nice…

[tame current]

Ever saw ded ppl replying

broo you need to create a ticket

[potent summit]

why the IOS app can read and the windows app no ?

[wind oar]

.

[hot steppe]

hello guys!

[twilit crown]

Why is all AI art just the same thing...

[distant aurora]

Fr

https://vxtwitter.com/zephyr_z9/status/1965155020761936269

Zephyr (@zephyr_z9)

be Dario Amodei, 2019
working at OpenAI
suddenly get a case of the super ethical fear feels about GPT-2
"We can't release this! The misalignment risk is too high! The paperclips! Think of the paperclips!"
quit in a huff of sanctimony to start my own AI company, Anthropic
our mission: to build a "safe" AI that is completely, utterly, and totally harmless
translation: so cucked it will actively refuse to tell you a joke if the punchline could be construed as micro-aggressive
need funding
look for people who are also terrified of a future where an AI might hurt someone's feelings
find my people: the Effective Altruism grifters
convince them the AI singularity is the real pandemic or something
get massive bags of cash from Sam Bankrun Fried, a guy who is definitely effective and altruistic
spend years and hundreds of millions of dollars building Claude
Miss the "GPT moment", cuz the AI wasn't aligned enough
realize we can't win, so we must cripple eve…

vxTwitter / fixvx · Sep 8, 2025, 8:47 PM

[keen gorge]

Heyy

[white plover]

did perplexity made changes to research mode with Social? i can't see sources and also the searches aren't from reddit in social mode..

[junior crest]

ma verrà mai aggiunta una modalità studio e flashcard?

but will a study mode and flashcards ever be added?

[green crystal]

basically the dude wants claude to behave like PC principal from south park lol

[versed sedge]

https://perplexity.ai/browser/claim/AVY6E4ZJN4

[halcyon oasis]

I updated my app and now the audio playback does not work on my perplexity app. I try to listen to my results, but it only gives me back a white noise sound. Is everyone experiencing this or what do I need to do to fix it?

[somber glen]

Thanks, taken.

[oak meteor]

No way he engraved it 😭😭

[green crystal]

@distant aurora oai deep research asking me if i need the report now or later lmao

they really need to update it

[scarlet spruce]

how do i make it so that perplexity gets rid of the attached image that ive already sent, like why does it stack up

[twilit crown]

Why is AI art just people making the same thing

[minor shard]

Lack of creativity maybe, same model

[distant aurora]

Bro was on lunch break.

[green crystal]

while 5 pro is way more thorough than deep research its being gimped by limited web calls lol....

and agent is being glitchy

[distant aurora]

I bet 5 thinking can rival research.

[green crystal]

around the 15 min mark it just pauses with no reason

[distant aurora]

Or maybe not idk.

I'll need to test with prompt for comparison.

[green crystal]

yeah on tasks that need a bit of focused search its better

if the goal is just mass info collection then research is better

[distant aurora]

They blocked auto thinking on my main account.

[green crystal]

lmao

[distant aurora]

It'll probably stay that way for a month.

It's already been a week I think.

[green crystal]

just use POE for thinking

[distant aurora]

I get only 300 credits per day.

So like one prompt.

[green crystal]

thats one rough region restriction

smh

[distant aurora]

Nano on poe does better than the "regenerate thinking" button for me.

[green crystal]

5 mini is pretty good too

seems like it sits between o4 mini and o3

[distant aurora]

Yeah.

Might be exactly O3 level but with search restrictions.

@languid star this is concerning https://www.vxinstagram.com/reel/DOfjs8ogRwZ

[green crystal]

on poe it can search lol

[young hollow]

Any New Ai Models

Specially Chinese Ones?

[young hollow]

FVCK Such Education Systems

[distant aurora]

@green crystal try this in custom instructions with gpt 5 thinking. https://sharetext.io/3dae8f42

[young hollow]

Afaik that inactives the openai system prompt for users pleasing

[green crystal]

Aren't such specific instructions going to force it to overthink?

[distant aurora]

Which you need for exhaustive research.

[green crystal]

The web calls are limited

[distant aurora]

However, my auto thinking is locked so it doesn't work on my og account.

[distant aurora]

Per prompt or per day/week?

[green crystal]

Sometimes during thinking you see it struggling. Trying to decide on what to spend Its last web call on

[green crystal]

Per prompt

[distant aurora]

What's the limit?

[green crystal]

Pretty high like nearing 15

[distant aurora]

Yeah. That should be enough.

[green crystal]

Other modes do way more

What's crappy is that 5 pro also shares the same limit

A model that greatly benefits from more input lol

[distant aurora]

Research does more?

[green crystal]

Yeah and agent

Though at this point only agent should be seriously used

Deep research hallucinates more than agent

Like allot more

[quaint ravine]

just ran Agent 7 times for identical PDF analysis and it got it right only 5/7

Research doesn't even use latest GPT's models afaik

[green crystal]

If it's not that large use 5 pro or thinking with more verifying

[quaint ravine]

dun have Pro, but friend who has it ran it for me and basically was barely 50/50

Grok 4 heavy also gets it wrong

[green crystal]

Using pro?

Interesting

[quaint ravine]

yes Pro got it wrong yep

well not all the time

[green crystal]

Is it large and complex?

[quaint ravine]

but about same % as GPT-5 Thinking

human who is aware of the field would easily solve it

it's very simple

but the PDF file is quite large

i know what the issue is tho

basically all AIs get confused by strikethrough text

or at least some degree of it while doing search/RAG

there's some clash

[jovial kindle]

@neon eagle

[green crystal]

That's new lol

[quaint ravine]

if u upload just a small amount

they seem generally fine

[green crystal]

If I upload strikethrough text as an image it gets it right

[quaint ravine]

yea

as i said

[green crystal]

OCR seems to be the failure point

[quaint ravine]

it must be something to do with my document being large

so it uses a bit more search/RAG

[green crystal]

That too yeah

[quaint ravine]

and then gets confused 50% of the time

[green crystal]

Agent is more free with file size

But it still won't accept my 5 million token file

[quaint ravine]

yep

Spaces will

[green crystal]

You mean project files?

[quaint ravine]

no PPLX Spaces

[green crystal]

Ah

[quaint ravine]

ChatGPT token limit is much less

for uploading files

pretty bad tbh

[green crystal]

I'm going to try with projects

[quaint ravine]

considering what PPLX and Google allow etc

[green crystal]

Maybe it will work there

[quaint ravine]

Projects same issue

[green crystal]

Oh

[distant aurora]

He uses documents beyond token limits so he's relying on rag.

And rag is shit.

[quaint ravine]

actually

for other documents it's 100% correct

that are even larger

the point is there are many variables

I've also shortened the document and it still goes wrong

[green crystal]

Interestingly notebookkm also doesn't accept my file

I thought it used rag

[quaint ravine]

that's surprising

maybe NotebookLM has a token limit of < 3 mil

it accepted my 2.5 million word PDF
so did Spaces

[winter fossil]

I'm exploring Perplexity. I expected it to be an AI search engine and a tool for verifying information, but I found it has many additional features, like workspaces, in-depth reports, experiments, etc. These seem to overlap with existing features in ChatGPT and the like, but after trying it a bit, the responses still feel more like those of an AI search engine rather than a content creator or researcher. Has Perplexity drifted away from its core?

[green crystal]

It was as a text file though

Maybe it's less restricted with PDF?

[quaint ravine]

doubtful

[green crystal]

Hmm

[quaint ravine]

try Gemini?

the platform

should accept

[toxic marlin]

Hey for the pro version do we get daily limits for research?

[green crystal]

Yeah it's a few million over the limit

[distant aurora]

https://chatgpt.com/share/68c5d788-0f14-8013-80cb-54efee231796

Seems higher.

[green crystal]

It's a chat log file though

[quaint ravine]

just worked for my 2.5 million word PDF

[distant aurora]

Result was still disappointing though.

[green crystal]

Guess I can't skip the work so I should chop up the date periods

[quaint ravine]

https://gemini.google.com/app

even free version works

[green crystal]

Not for 5 + million

[quaint ravine]

Spaces should work

that's interesting that Google has poorer limits than PPLX in that regard lol

[novel salmon]

https://perplexity.ai/browser/claim/TG783S0HIC

[distant aurora]

God bless you.

[young hollow]

Hi Billy

[dry kestrel]

#💬│general message

[young hollow]

First check the question bro

[solemn cave]

Hmm

[winter fossil]

What problems are you overcoming?

[young hollow]

The Intelligence Problem

[dry kestrel]

They were talking about Perplexity I thought?

Wdym

[distant aurora]

Hey

@green crystal Lil bro will ask for his rights soon.

[ripe tulip]

they were, but you linked a how to get the pro role in discord, for a question about research limits?

[young hollow]

Why don't you name Yourself BillyBilly?

[young hollow]

He is asking for Perplexity Research Limits not How to Get The Pro Role...

[toxic marlin]

Yeah basically this .....

[dry kestrel]

#💬│general message

[dry kestrel]

Linked the wrong message, not sure why that was on my clipboard

Oops

[toxic marlin]

But the thing is I haven't made a single mssg yet and still it shows 11 queries left...

[serene sand]

which country?

[dry kestrel]

Should not be the case, please contact [e-mail]

[toxic marlin]

India

[toxic marlin]

I tried but they said the same stuff.....

[dry kestrel]

It could just be a temporary issue, in your email, specify "I want human support"

[toxic marlin]

👍

[green crystal]

Talkative clanker
Needs to be spanked

Also

@quaint ravine project works

Gonna test it's depth of analysis later

[quaint ravine]

Maybe ChatGPT Pro has higher token than ChatGPT Plus - I just tested my 2.5 million word PDF and it failed. Maybe text files work - will try to convert the PDF to .txt and update
Nope didn't work either

[hallow jewel]

hi yall

[young hollow]

Ahhh i see

[serene sand]

sonnet/web off

w/ low temp, still dreamy dreamer

[winter fossil]

What are you hoping to achieve with Perplexity?

[quaint ravine]

We were talking about ChatGPT platform

PPLX Spaces works no problem with millions of tokens uploaded

[winter fossil]

Sorry, my mistake — I misunderstood; I thought you were talking about Perplexity’s "Research" feature

[lyric badge]

It's always very vague with perplexity

[quaint ravine]

that was ChatGPT

just edited my comment to clarify (we had a conversation earlier about this)

[rustic hull]

So does someone wanna give me a quick rundown on the diff between my perplexity pro subscription, my chat gpt pro subscription, and my google ai pro subscription? obviously i know the obvious like perplexity has comet as well as all the top models from each company built into it. What i am confused about is if perplexity has gemini 2.5 pro and chat gpt 5 and all the best llm then why is it not just as capabale in the exact same areas that those models are?

would perplexity just be all around bettewr becasue it has all the mdoels in one? or are the hamstrung?\

[winter fossil]

Can't you get the answer you want by asking Perplexity directly?

[rustic hull]

I get different answers from each of them

so i was asking the earthlings

[serene sand]

#💬│general message

deep dive ➤ #💬│general message

[rustic hull]

I honestly had not heard of perplexity until a couple days ago, but im facsinated now. hence my subscription. I love how comet seems to be built on chromium or the same browser base as chrome

ty sir

[winter fossil]

I'm more curious about how Perplexity verifies the information it finds. In my country there are a lot of media outlets spreading false information, and I often have to fact-check two opposing accounts to roughly determine which one is telling the truth

Basically, the first page of Google results I find are always news released by media outlets — they may not be false, but they're always slanted toward one side

[ripe tulip]

I guess it can't in full probably just weighs certain more trust worthy sources higher then others when deciding what info to include

[quaint ravine]

guys did the source search and model selector just swap places?

[winter fossil]

Yes, they really swapped places

[quaint ravine]

hopefully bug fixing 🙂

[serene sand]

https://www.perplexity.ai/search/i-m-more-curious-about-how-per-SNS_jlnjSlO_ySZ31AopYw

[distant aurora]

In Gemini you get 1m context, per day veo 3 limits, notebook lm, and their own 20 deep research per day which tends to be much better.

[distant aurora]

Chatgpt pro or plus?

[brave phoenix]

Is there any spcial discount for students if they want to get perplexity max?

[distant aurora]

They ain't giving discounts for a 200 dollar sub 💀

And I don't think you should be even spending that much money as a student.

[ripe tulip]

Find it funny how many people throw large amounts of money at AI without fully checking limits etc of shit first

[distant aurora]

Fr.

[winter fossil]

I don't notice a significant difference between Perplexity Pro and Perplexity Free right now — maybe my everyday needs don't require that much

[distant aurora]

💀

[ripe tulip]

Saw one person spend $700 on vibe coding just to fail to implement stripe payments. Because they failed to realise you still gotta know what your doing to vibe code and more money isn't a solution

[distant aurora]

Yeah probably.

[winter fossil]

Most people wouldn't understand the meanings of those AI-specific technical terms..

[distant aurora]

Free tier model is too dumb for science-math and complex queries.

[dry kestrel]

No and please don't spam the same question in every channel. There is a student plan though which gives you access to study mode and Comet

[distant aurora]

Good enough for quick searches though.

If you are using free tier then you might as well use qwen or kimi.

Qwen offers strong reasoning models for free.

[winter fossil]

Few AI solutions emphasize use cases, allowing consumers to understand their needs more intuitively

[distant aurora]

Qwen, kimi and z.ai are probably the best free tier rn.

Also, a general advice

Please do not spend 100-200 dollars on anything as a student unless your field needs to use/spam deep research like features.

[winter fossil]

Basically, Perplexity’s features are fully accessible; if your usage exceeds the current plan, Perplexity will prompt you to subscribe

My understanding shouldn't be wrong, right?

[distant aurora]

For max?

It is free. I don't think there's any limit on number of queries.

[winter fossil]

I mean free

[distant aurora]

Limits are on uploads,model selector, and context size.

And fancy stuff like image generation.

[quaint ravine]

does Max change uploads and context size?

[distant aurora]

There's no clear information about context anywhere. Kwak said that reasoning models use 128k.

[quaint ravine]

yea, but that's regardless of Pro or Max?

[winter fossil]

Will Perplexity provide a warning when demand exceeds the current plan? I think this is needed

[distant aurora]

Yeah

#1345432443074646158 message

[quaint ravine]

yea

[distant aurora]

#1345432443074646158 message

[quaint ravine]

was interpreting ur previous statement as if Max changed something

for those parameters

[distant aurora]

Nah. You get O3 pro and Opus.

[quaint ravine]

indeed

[distant aurora]

And no limits on queries.

[quaint ravine]

and less limits on images/vids

[distant aurora]

And you get to pick model for research and labs I think.

[quaint ravine]

no dun think taht's the case anymore unfortunately

it was briefly a possibility when Max first got announced

[dry kestrel]

It will tell you that you've reached your rate limit. This is very unlikely to happen on the Pro plan. Learn more about rate limits here. #💬│general message

[dry kestrel]

Higher video and image generation, better agentic model on Comet

[quaint ravine]

it's not amazing limits still for video iirc tho, but then PPLX isn't meant for that sorta stuff really

already amazing it allows stuff like that

[distant aurora]

@dry kestrel okay explain this part "research and labs also run on more powerful models" #💬│general message

[wise nova]

Mobile Web UI was updated and it get worse , new citations

[quaint ravine]

do u remember the exact link? I suspect they have updated that page now

[dry kestrel]

Afaik you can't choose the model yourself.

[distant aurora]

I could be wrong.

[dry kestrel]

It never states being able to choose the model yourself either on the FAQ and in your screenshot.

[quaint ravine]

actually looks the same

yea the ability to choose model was intiially a thing tho...there are even some reddit screenshots explcitily describing that in Max plan descriptions

but was removed pretty quickly and I think when Max officially released, it wasn't a thing

https://www.reddit.com/r/perplexity_ai/comments/1lo7mvp/perplexity_has_launched_a_new_tier_called/
yea I remember those 3 models made it into the GUI under Research/Labs (but wasn't active yet)

[junior crest]

.?

[young hollow]

Gn peeps

[distant aurora]

Gn bro

[young hollow]

You should sleep too bro

[distant aurora]

I'm going through things I do not know how to handle.

[winter fossil]

Good night

[quaint ravine]

ask AI?

[young hollow]

No matter the problems understanding is the only solution

But if is tough is hard to even think clearly

[young hollow]

[distant aurora]

Nah. There's things beyond ai. Gathering and feeding accurate data. Something it doesn't work well for in certain life scenarios.

I tried what I could.

[young hollow]

And what exactly is your problem?

And if you really can't handle it if is out of your control you should just let it go perhaps

[strange atlas]

IS comet free now?! 😭

[distant aurora]

Job hunt.

[young hollow]

FVCK

[distant aurora]

Was always free.

You want an invite?

[young hollow]

That's why im just living in poverty with family

[strange atlas]

https://tenor.com/view/what-surprised-gif-25739412

[young hollow]

Never have to care about education or jobs

[strange atlas]

I have one already

[young hollow]

Which will be taken anyways

[strange atlas]

but whats the purpose of getting max anyways?

if comet is free

[distant aurora]

Okay so why are you asking?

[distant aurora]

Higher rate limits I'm assuming.

[strange atlas]

Im asking if its free dude

[strange atlas]

and the benefit of getting an invite version?

[young hollow]

There are some limitations for Free afaik

[distant aurora]

Yes comet is free. You should have known this by know if you had installed and run it.

[strange atlas]

i gotten invited lmao

[distant aurora]

Nothing.

Early access.

It isn't available for downloading without invite yet.

[young hollow]

Well im sorry for you pain i hope you find a job or something

[strange atlas]

to beta features or just old?

[young hollow]

Welp gn now gotta go

[distant aurora]

Americans can buy invite via pro sub.

[young hollow]

[winter fossil]

Speaking of which, Comet doesn't even have device sync — isn't that the first thing it should have thought of?

[dry kestrel]

Higher rate limits, better models, instant access to Comet, better agent in Comet

[strange atlas]

ah I see

[strange atlas]

thanks for asking

[waxen raven]

perplexity is horrible

[distant aurora]

He meant informing.

[waxen raven]

can someone explain if im missing something

chatgpt is so much better

why is gpt 5 on perplexity so bad

is there a setting i need to change

[dry kestrel]

Make sure you are selecting GPT-5 Thinking

[waxen raven]

ok i will try that

[distant aurora]

Yeah.

[dry kestrel]

It should be the same underlying model that is used on ChatGPT.com

[quaint ravine]

should be fairly similar for normal/many queries, except PPLX has a different way of searching/RAG compared to GPT's own platform

GPT5 on PPLX is basically = GPT Instant

while GPT-5 Thinking should be same

[green crystal]

That's weird

[quaint ravine]

nice that worked for your 5 million token file 🙂

[green crystal]

Yeah I just checked and it was able to parse and interpret all messages though I gave it some space by only asking it for a certain period

It basically was able to do that by utilizing python lol

Something would probably break if it would try to read it directly

Interestingly, though it only looked at a small part of the text file and then assessed the message structure and created regex for it

[quaint ravine]

yep PPLX Spaces is quite impressive too for huge text

[green crystal]

I'm going to try agent later

[quaint ravine]

what's your word count btw?

for that file

[green crystal]

Not sure but Gemini in studio showed me almost 8 million tokens

And that's after cleaning

Before cleaning it was 15 million

[quaint ravine]

wow

[quaint ravine]

I can't even upload my 2.5 mil word file into AI studio

says failed to count tokens lol

ah converting the PDF to text file works in AI studio -

[quaint ravine]

"All text and document files uploaded to a GPT or to a ChatGPT conversation are capped at 2M tokens per file. This limitation does not apply to spreadsheets."

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

what exactly is the file type you uploaded?

okay I managed to upload it in CSV format lol

nearly 4 million tokens

[karmic isle]

hi

[wise nova]

After today's UI updates, the code output was broken.
Claude 4 Thinking

[fast meteor]

using gpt5 by any chance? try other models

[wise nova]

Regardless of the model,

[fast meteor]

pls make a #1140622008086970420 and make the thread sharable there

[stoic crater]

I can bet OpenAI has just enabled GPT 5 on their advanced voice mode

It sounds better

[molten dragon]

lol why

why only voice

[wise nova]

Already refined the thread, it's gone.
I work with xml and md output, and pplx just awful with any code output.
Today UI update broke output so much

[tacit locust]

hello

[distant aurora]

@green crystal https://www.vxinstagram.com/p/DOjWG-jCIhJ/

[wise nova]

If the situation with the code output does not change, then I will be forced to switch to native ChatGPT sub

[hidden lantern]

hi

[green crystal]

[green crystal]

a massive whatsapp group chat log lol

as a text file

it still took it despite the much higher token count

but the model resorts to purely using python to snip out stuff

[quaint ravine]

yea .txt file right? interesting

I can upload 4 million tokens of CSV format, but not .txt format

[green crystal]

ok agent is working again

so you can upload it to drive

connect that to chatgpt and just add it through google drive

then you can use agent to work with it

[wise nova]

Gemini much better handles code blocks, clean and accurate 👍 shame to pplx

[dry kestrel]

Please make a bug report containing how to reproduce the issue, the prompts you used, the thread in which you experienced the issue, device, platform, etc.

[modern sinew]

There isn’t actual problem with code blocks in perplexity

The only issue which happens is with gpt 5 instead of returning in code snippet it returns without it

Another feature you could add is that the copy button for code snippet should be sticky so even if we scroll we can still see the copy, this way we wont have to go back to the top where currently copy button lies

Team isnt adding this from a long time

They are adding more bugs then adding useful ux features

Thats why i dont use perplexity as my main ai tool anymore

Gemini is better

[dry kestrel]

The non reasoning variant of GPT-5 really doesn't like to adhere to the system prompt. I've had this issue myself a ton while trying out the API. Also the user wasn't specifically talking about GPT-5

[dry kestrel]

Also if you have any product suggestions please post them here #1358914210339360798 or #1111765634267742408

[dry kestrel]

Gemini trains and huaman reviewers views your conversations even on the paid plans, and the only way to disable this is to turn off conversation history.

[modern sinew]

I am no research engineer who will spend 300 requests.

Perplexity is just bugging more then being useful

Rate limits doesn’t matter to me both platforms provide a substantial amount of requests, ux actually matters perplexity is buggy for my usecase.

Will wait for team to fix the buggies and then will switch

[dry kestrel]

Please post bugs in #1140622008086970420 . 100 requests a day is quite low for the price of the Google AI Pro subscription. I found that Gemini models tended to refuse requests more often compared to on Perplexity.

[gleaming summit]

hello

[indigo osprey]

Hello

[golden palm]

Are quantum computers better then ai?

[quaint ravine]

interesting, just tried again, and now ChatGPT allows my 4 million token text file (which I converted from PDF) to be uploaded both in normal Chat and Projects 🙂

maybe I wasn't able to before due to traffic etc lol

[green crystal]

and uh

only models themselves can read it

agent cant

[nimble surge]

yes.

[golden palm]

Damn

[nimble surge]

and ai use quantum computers

[golden palm]

Gpt quantum 7

🐧

[quaint ravine]

yes in Projects attached file, Agent seems to fail...lemme test in normal chat

yep, Agent works with normal chat attachments

with attached file

seems Agent struggles to find the files attached in Projects

[green crystal]

Yep

[slate flume]

the moderation team in this server is so awesome, after almost over 12 hours, a pedo who got reported by multiple users is still not banned and mods just straight up ignoring reports from multiple users

[quaint ravine]

I prefer my PPLX Spaces answer for this so far tbh

with GPT-5 Thinking

[wise nova]

One big problem of Perplexity is thr focus only on searching/surfing instead of multitask and multi domain using,

[quaint ravine]

one cool thing Agent did though was reference the exact words/sentences for some points

altho looks inaccurate/not optimal tbh

[modern sinew]

What do you mean by "multitask" and "multi domain"

It goes* over my head

[wise nova]

Any code interactions, coding, analysis tools, math, logic engineering tasks, etc. But pplx offer only chill surfing instead applied tasks. Compare with ChatGPT capabilities

[tame current]

CodingBoyBlah

[vivid ferry]

the efficient research is peak for up to date API references.
And switching between gpt5, sonnet, and gemini is peak, itd suck for me to use a single provider as their models arent equal

so its just the UI thats behind, otherwise I think its still better than those others

[brisk igloo]

hello

[candid otter]

why are we even here if its isnt a good app

[languid star]

fr

we should leave

starting with you sir

[modern sinew]

I will leave too if hammy is leaving

[slate flume]

Just get chatgpt plus

[languid star]

https://tenor.com/view/homelander-the-boys-speechless-sad-give-me-a-minute-gif-5336379821034774682

[nimble surge]

you know the browser is not an llm

it has api access but thats it

[dark kettle]

Where did billy bones go

@distant aurora

Omlandh is here

[languid star]

https://tenor.com/view/sleep-gif-12944899486400191725

[modern sinew]

Billy bones is sleeping

[dark kettle]

Oh

[languid vale]

Hello

Hello Perplexity Team,

I want to express my appreciation for your platform, but I’ve run into a persistent issue when using Perplexity on my Android browser. Here’s what’s happening:

Problem:
When I write a prompt in the main text box and then switch to another app or browser tab, the draft prompt disappears when I return. Unlike other platforms, my work isn’t saved unless I send the prompt—which isn’t practical for longer, multi-step prompts.

Impact on My Workflow:
As someone who regularly drafts complex prompts for AI experiments, this is a major inconvenience. I might need to research or refer to something else while composing, and I expect the draft to be there when I return. ChatGPT, for example, preserves the draft for days—it only clears when I send or delete. This is especially important on mobile, where I often have to switch contexts.

Why This Matters:
This change would save time and avoid frustration for all mobile users. For users like me who build prompts iteratively, draft persistence is essential.

Suggested Improvement:
Please consider preserving the draft in the text box—whether that means saving it locally, syncing it with accounts, or just keeping it active in the session. A “Save Draft” or “Recover Prompt” feature would be even better.

Thank you for considering this feedback. Improving this would make Perplexity more reliable and versatile for mobile prompt engineers like me.

Best regards,
[]

[shrewd thistle]

vote this guy

[languid vale]

You also faced the same issue

Actually I have been using perplexity for a while now and I am surprised by how well the perplexity sonar works. In fact many times I prefer its answer over the chat GPT-5 many times.

[languid vale]

I posted it in the feedback channel as well. Vote this feature there also to catch the attention of the developers.

[shrewd thistle]

itts great for internet search etc

guys is claude 4.0 thinking sonnet is the only model with some braincells.

i had7 chapter pdf (20k tokens) and i needed alll the fformula extracted from them, i first gave it to research mode it missed 6 essential formula . then i gave to gpt 5 thinking inside perplexity , it also missed the same formula, then to gemini 2.5 prro inside perplexity same shit, then i gave to gemini 2.5 pro at ai studio and completely perrfect all formula it gave, then i gave to cllaude 4.0 thinkinng sonneet inside perplexity and it also gave whole and in much better way than aistudio.

[languid vale]

Thanks for sharing.

[shrewd thistle]

can you gimme your feedback message link

[languid vale]

How?

[carmine sandal]

Dawg Why the study mode isn't displaying?

Study mode feature

been pissed bout this.

anyone know bout this?

[haughty vessel]

hi

[carmine sandal]

I have mid sem Tommorow, and using my insitiuite mail but study mode feature isn't displaying

im bout get fucked rn

[crystal snow]

guys to make ai chatbot like chatgpt or perplexity which skill we have to learn ?

[shrewd thistle]

USE NOTE BOOOK LLM

[shrewd thistle]

brrother you just cant

you can use their api

to make your own chatbot platform

like name it chatai, but in background its using chatgpt etc

[young hollow]

Yeah You Need to Pay for API

Chatbot niche is already so much saturated anyways

[crystal snow]

but i have to make my own

[young hollow]

To make your own you would need to train a llm and have compute power to run it

[shrewd thistle]

is research mode good for formula extraction from 7 pdfs worth 20k token @young hollow

[young hollow]

Idk tbh but i think it should work

[languid star]

tbh ngl imo it should work fr

[young hollow]

Yeah because is just extraction

Although idk the context window limit for research

But even for non reasoning models is 32k

[shrewd thistle]

But it fking doesnt literally the onnly ai which actually extracted all the formula fromm 20k token pdfs was claude thinking 4.0 sonnet not even the research mode could do it man, though yeah ai studio gemini 2.5 pro did that too. but in perplexity only claude thinking 4.0 sonnet

other model missed 8 formullas

8 . I GOT A FKING EXAM tmr AND ITS GONNA make me fail if i ask any other ai model seriously

[young hollow]

Then use claude or google studio

[shrewd thistle]

man but thats the thing i really used only perplexity due to how great it was man, but nnow ik imma stick with claude sonnet 4.0 thinnking in perplexity its the best model for such stuff

i used gpt 5 thinking etc nothing worked

just claude 4 sonnet thinking inside perplexity answered

[young hollow]

I see

Well just use whatever work

You can even try the chinese ai models

[quaint ravine]

nothing worked like how - seems like over-generalisation fallacy?

ime, no AI is perfect for PDF extraction/interpretation

but ChatGPT's platform might be one of the best

[shrewd thistle]

i mean its not my first time seeing chatgpt act so rubbish with pdf''s

[quaint ravine]

talking more about the ChatGPT platform tbh

[shrewd thistle]

earlier durinng holidays i uploaded a huge pdf and asked it to solve the last page exercise

[quaint ravine]

not PPLX + GPT-5

[shrewd thistle]

it couldnt even find that forget about solving

[shrewd thistle]

yeah i am talking abou chatgpt too

that time i didnt had pplx

[quaint ravine]

how long was that

i believe GPT-5 Thinking has improved things a lot even on ChatGPT platform

for things like this

I'm actually finding Claude Opus 4.1 to be one of the worst for accuracy of PDF analysis...but it's all small sample size of course

you need millions of tests to achieve statistically significant differences imo

[shrewd thistle]

i am using 4.0 sonnet thinking

not opus

qwen also answered perfectly

[quaint ravine]

Opus 4.1 is supposed to be the strongest Anthroopic model?

[shrewd thistle]

yeah thats true

lmao but i cant use it

[quaint ravine]

but yes, I agree...in my small sample size, I found Sonnet 4 Thinking better lol

[shrewd thistle]

huh bro damn

[quaint ravine]

yep