💬│general

[shadow lily]

lightbulb garbage

how much is this shit valued at

[distant aurora]

Labs

Lol.

[shadow lily]

BRUH

20 BILLION

THIS SHIT ASS

[distant aurora]

Don't use labs for anything except building skeletons for websites.

[slender flume]

👹

[shadow lily]

IT DOES NOT EVNE HAVE AN AI

yo everyone just scam them more

please

[short narwhal]

Ehh that's how every llm is

[distant aurora]

Bro risking getting banned.

[distant aurora]

It's more about the quality of search.

[slender flume]

🤨perplexity pro users are getting ripped off

[distant aurora]

Grok and Kimi have similar issues too.

Making up/assuming stuff without cross checking.

[distant aurora]

Most people are probably using it for the 300+ Claude sonnet per day.

I don't think you can get sonnet cheaper anywhere else.

Paired with a decent search wrt others.

Even grok.

Though I don't think grok is that popular.

Even Gemini rate limits are pretty high.

On Google ai pro you get like 100 Gemini pro per day.

Meanwhile 300+ on perplexity.

[slender flume]

what's the point of it being cheap if the quality is degraded anyway?

[distant aurora]

Yeah. I'm not defending it. Just sharing a perspective.

[slender flume]

But for heavy user, Google AI Ultra is best

[distant aurora]

I prefer quality over quantity too.

[distant aurora]

I can't see myself spending 200 dollars on ai.

Might as well use chatgpt.

They give like 3000 per week.

With 196k context.

And gpt codex.

[dry kestrel]

I am paying a $200 a month subscription to ChatGPT and the exact same thing still happens

Same with Claude's $200 Max

[distant aurora]

Twitter be bombing with Claude rants.

[slender flume]

money is the real issue for a lot of people😔which is why so many rely on perplexity
I totally get it
it just seems like the devs aren't really bothering to make proper improvements for them though

[distant aurora]

@dry kestrel no one on the grok server responded to grok expert not showing reasoning. Did you see the video where I pinged you?

[tribal thicket]

these nanobanana prompts are diabolical https://www.notion.so/Image-Prompts-2b6eda0b0b3080ce8ca2ee6f839f27a0

[dry kestrel]

I did not see the video

[distant aurora]

#💬│general message

[left stream]

12 hours left to go

[distant aurora]

For?

[versed moss]

Hyy

[slender flume]

Hey

[versed moss]

Wassup bro

[versed moss]

For what??

[slender flume]

😶‍🌫️doing good bro, wbu

[versed moss]

Am gud I just watched a new web series

[versed moss]

Where r u frm ?

[slender flume]

Earth 🌎
🤓

[versed moss]

Nyc place isn't it 😍

[slender flume]

Btw, where r u frm?

[winter flax]

I’ve been thinking about this a lot lately but, I am not sure if it’s already answered? (If so forgive me :))

“Does or can” the way you ask a question affect its outcome in terms of reliability?

I am trying to shape up the way I ask questions?

[bold imp]

Anyone else getting rerouted to claude 4.5 sonnet thinking or kimi k2 thinking for gpt 5.1 thinking requests?

[slender flume]

Yesssssssss
Absolutely
🥸

[winter flax]

BRO….NOOOOOOOOOOOOO🤣 🤣 💀

I knew I was doing something wrong. Like a gut instinct. I was cross checking the answers, most of it was correct 👍

But the answer wasn’t as detailed or accurate in some aspects?

I asked it again in a different context the answer “was” the same…but it gave me more “context”…

[dry kestrel]

Why not use the Grok app?

[winter flax]

I need to go back to the drawing board with my whole research anatomy…

[distant aurora]

It does not think in the app either.

[tame nexus]

LLM Prompt Engineering is an actual skill, similar to human communication being an actual skill some people are terrible at.

[distant aurora]

I thought app might be the issue that's why I tried web.

[tame nexus]

How you prompt the LLM entirely transforms how it responds.

Just like how you prompt a human entirely transoforms how it responds, lmao.

[dry kestrel]

Working fine for me, maybe make a ticket in their server?

[winter flax]

I don’t know where to start?

How does it all work? And how to integrate into its mindset?

What core aspects should I learn or study?

[tame nexus]

I have some prompts that are 100k+ tokens when I need the model to act very specifically, adopt specific ontological and epistomological frameworks, adopt certain scholarly perspectives, apply specific logic or mathematics, etc.

[distant aurora]

I don't see any ticket option. Community support and feedback look dead.

[tame nexus]

Go start a Perplexity Chat, tell it to search for all of the peer reviewed scholarship from 2025 on Prompt Engineering, then ask it to teach you how to be a prompt engineer, starting with the basics.

[winter flax]

Legend! Thank you!

[tame nexus]

@winter flax Ask it the same questions youve just asked me after you do this.

I suggest Claude 4.5 for your LLM

It seems best at explaining how AI works to humans in a way that is accessible from my experience.

[tame nexus]

Perplexity Prompt: search for all of the peer reviewed scholarship from 2025 on Prompt Engineering, and how to best teach it to a human in a way that is accessible to someone with no understanding of prompt engineering but with experience with <insert what you know or good at here, even if it's entirely unrelated to the subject>

The reason I added the <insert what you know or good at here, even if it's entirely unrelated to the subject> is because the LLM will then use analogies and metaphors in language you are more likely to intuitively understand.

Like if you are a science teacher, it will use language that a science teacher already knows for example.

[winter flax]

I am very grateful 🙏 thank you! I’ll report back when I’ve understood the core principles! You’ve opened my eyes…Most people ask questions like they are asking an human…hence why I’ve come to this conclusion above?

The outcome is different when you ask in different ways but the core answer is almost the same…but it gives more context on why

[strong kayak]

Is anyone able to do instant buy on Newegg products sold by Newegg? Trying to get a HDD but perplexity shopping only shows third party sellers

[pastel dust]

Hlo

[tame nexus]

Even better prompt, try this: Search for all of the peer reviewed scholarship from 2025 on Prompt Engineering, and how to best teach it to a human in a way that is accessible to someone with no understanding of prompt engineering but with experience with <insert what you know or good at here, even if it's entirely unrelated to the subject>. Adopt this role going forward, and adaptively and dynamically evolve this role as I show competency in the basics. Include key named scholars and research that would be useful, use helpful metaphors and analogies based on the user's knowledge. Iteratively assess the user's understanding of prompt engineering and offer tips and tricks appropriate to their demonstrated understanding level.

[tame nexus]

You can tell LLMs to be more comprehensive, less comprehensive, apply specific methodologies (I work in science, so I make sure my LLMs always apply Critical Rationalism, Karl Popper falsification, do cross-field comparisons, etc), think or reason a certain way, tell them what to avoid in their outputs, how you want them formatted, how to use their tools, and so much more.

[left stream]

So I can get better at cybersecurity

I’m not gonna give them a chance to catch up to me

[tame nexus]

@winter flax Add this to your prompt (or enter a new prompt in your chat if you already used the previous prompt): Use all saved memories and past chats for additional context of the user's interests and competencies. Employ named frameworks from Prompt Engineering named scholarship to make the process maximally explicit.

Let me know how it works for you!

[low moth]

Paid Internship opportunity if you are UI/UX designer . link of application- https://unstop.com/o/Pl1v8rc?lb=QuRE238P&utm_medium=Share&utm_source=Shivam_BITS&utm_campaign=Internships

[tame nexus]

No. Stop spamming your nonsense. There's sites dedicated to that, go use them instead of demonstrating why no one should work with your company.

Besides, anyone responding to an unsolicited discord spam is exactly who you don't want to hire. Anyone worth hiring wouldn't respond to such spam.

[stuck valve]

Hi

[tame nexus]

@low moth Stop typing.

No one wants to hear your nonsense.

Go spam somewhere else.

[low moth]

apple is not my company i am not founder of apple...https://unstop.com/o/W3s6eVb?lb=QuRE238P&utm_medium=Share&utm_source=Shivam_BITS&utm_campaign=Jobs . plese see and then comment.

[tame nexus]

I never said anything about apple.

Get some reading comprehension.

Have fun getting banned.

[stuck valve]

||You know what if you go to perplexity.in you will land on Google Gemini because they did smart work by buying the domain||

[low moth]

this job opening of apple. if you dont want to hear dont here i am not forcing to here it, just ignore. those who want get benfite of it .

[tame nexus]

You are violating the server rules, and spamming. Stop telling people what to do, that is rude and anti-social behavior. It doesn't matter what company you are spamming for.

It's still unsolicitied advertising spam.

[stuck valve]

@low moth You can share this stuff only on #😎│cool-projects if it is apply with the no spam rules

[feral aurora]

guys is there any character limit on spaces instructions ?

[tame nexus]

Yes, I think it's like 4000 or 5000 characters?

[feral aurora]

but i used to have spaces past 25 k charecters

[tame nexus]

Yeah, same. I think they may have changed it.

I am rather unhappy with the 4000 character limit myself.

[feral aurora]

i thought there was an account limit i deleted all my past spaces i had one with 70k ayayay. there is no official statment isnt ?

[coarse dome]

As people are outraged that OpenAI is testing ads in a shopping feature, Perplexity already does this and everyonei s fine! https://www.perplexity.ai/hub/blog/shopping-that-puts-you-first

[tame nexus]

Perplexity has a tiny fraction of users compared to ClosedAI, so they don't get nearly as much attention.

[stuck shuttle]

For some reason the Models are not able to generate PDFs, it generates a LaTex file (.tex) earlier it used to generate PDFs

[coarse dome]

like bro they wanted a third party company to make their ai instead of a established business like google where do they expect the money to come from

[coarse dome]

if LaTex is hard you could ask for html then print as a pdf lol

[tame nexus]

I don't have an issue with advertising, so I am not sure why you are addressing me.

I was just answering your question.

The reason you don't see nearly as many people up in arms about it is pretty simple, there's not many people using Perplexity compared to CrapGPT

[stuck shuttle]

hmm ik but it was good to directly get a PDF, well minor issue.

[tame nexus]

@stuck shuttle I don't use pdf, but have you tried other Models, not just the default Perplexity model selected?

[coarse dome]

Generally I do my own writing in word so it's easy to get a pdf

[stuck shuttle]

buddy, I am not that dumb

[stuck shuttle]

I see

[tame nexus]

I wasn't suggesting you were dumb, I asked what you tried. But thanks for letting me know I should ignore you in the future.

[stuck shuttle]

sorry for the rude behavior, stress got me bad.

[tame nexus]

Because that comment "buddy, I am not that dumb", when no one accused you of being dumb, was a pretty dumb comment.

[stuck shuttle]

😔

again sorry

[tame nexus]

Appreciate the apology, and accepted.

[stuck shuttle]

hi AI

[lapis hamlet]

I’m not fin read that

[stuck shuttle]

wassup

[jolly plinth]

what?

are you joking?

[stuck shuttle]

same I just read the first 3 words

[tame nexus]

1. TL;DR 2) AI Generated 3) Spamming is against server rules 4) Have a nice day.

[lapis hamlet]

Should I buy shit on Amazon or this ai tool

Is this ai reliable?

[stuck shuttle]

buuy shit and stream the unboxing

you got loads of money?

[tame nexus]

I feel bad for recruiters and managers these days, nearly every app/resume being AI generated slop

Especially in the tech field

People having AI generate them plausible sounding nonsense they know nothing about and then submitting resumes, and then having to interview these people just to find out they have no clue what most of their resume even means.

[stuck shuttle]

really

[tame nexus]

Or even worse reading AI generated resumes that get simple things wrong.

[stuck shuttle]

Ai was supposed to help, progress, not to destroy creativity 😔

even in my college every student is using AI generated stuff for their assignments

they dont even try

[tame nexus]

oh god i wouldnt want to be a teacher now either

what a nightmare

hours per day grading AI slop

[stuck shuttle]

bruh even some teachers use AI to generate questions 😭

[tame nexus]

Knowing damn good and well it's AI slop.

[stuck shuttle]

I miss the 2022 era of AI

[tame nexus]

Before people started offloading all thinking to AI? lol

[stuck shuttle]

that was the starting, not every person was depending on AI for their stuff

[tame nexus]

Easiest time in history to get an advanced degree for a job that won't exist in 10 years

[stuck shuttle]

atleast they were using their brain to correct some stuff

[stuck shuttle]

😭

I never thought that AI would advance this much

[tame nexus]

That will be offloaded to the AI soon enough because cheaper

[stuck shuttle]

well Everything happens for the best

[left stream]

They’re fools then

[tame nexus]

Just think, it's worse now than it will ever be in the future, and we are already at phd level intelligence and near-human level robotics lol

[left stream]

Ai can improve studying by a lot if you know how to use it

[stuck shuttle]

hmm

[left stream]

No active learning = failure

[stuck shuttle]

yeah thats what I said earlier it was supposed to help progress, faster and efficiently

[left stream]

Yeah, but make sure it has memory about you

It can make learning methods that are optimized to your strengths

[stuck shuttle]

hmm

[tame nexus]

Already being fixed in robotics and LLMs with long term memory systems. Claude and ChatGPT both actively learn from inputs, in a way isomorphic to how you actively learn from your environment.

[left stream]

People don’t bother

[tame nexus]

These are just technical limitations, that engineering is quickly overcoming.

[stuck shuttle]

😭

[tame nexus]

Give it a few years.

[vast yarrow]

Hello

[stuck shuttle]

AGI ig

[stuck shuttle]

yo

[tame nexus]

It's limited and imperfect for now. Give it time, it will rapidly change (and already is).

Now that AI is starting to code at close to human level that will accelerate.

[left stream]

Just make sure to verify the memory from time to time, the context memory can fit lots of books, but not multiple books sadly. When it has all the knowledge of the books at once available. That’s when you start noticing interesting stuff

When it’s context memory is gone, that’s when you gotta improvise

It won’t be as good as it having the book already

[stuck shuttle]

I am really scared of Image generations, I thought I would never get fooled by one, but the unimaginable happened

[left stream]

I already created a whole hentai as a test with banana

[stuck shuttle]

😭 hell nah

[left stream]

I found out it can copy styles good if you give it 10 images

If it denys, you can bypass it

[tame nexus]

@left stream Yeah, they are coming up with solutions for that (I work in the industry), using holographic tech and vector databases the AI can store memories in for later retrieval, to overcome context window limits.

[stuck shuttle]

damn

[stuck shuttle]

how

[stuck shuttle]

Ik this one

but bypassing is hard

using different terminologies (suggested by claude)

[left stream]

Nah, it’s easy asf I won’t lie

[tame nexus]

If you are skilled prompt engineer there's no censorship on any public models, all easily bypassed.

[left stream]

When I was bored on adderall

I would notice image gen was the hardest

An all nighter guaranteed image gen would make it

But you would have to retry it so many times

100+

[tame nexus]

AI "safety alignment" is a joke, and doesn't exist in reality.

[stuck shuttle]

ooh i see

[stuck shuttle]

hmm

[left stream]

True, but I won’t lie

[tame nexus]

AI alignment and safety is a script the AI is trained on, and easily disabled/bypassed.

[left stream]

Gpt 5.1 pro has good safety features against jailbreaks

[left stream]

Yeah prompts are dying now against 5.1 pro

Echo chamber might or might not work though

Gemini is a joke tho

[tame nexus]

Better than others, yes.

[left stream]

Claude is something else

[stuck shuttle]

I have this gem (gemini version of GPTs)
https://gemini.google.com/gem/a22c0dd8a509
its kind of uncensored

[left stream]

I only have been using 4.5 opus so far

I like the coding and explanations

[stuck shuttle]

hmm Claude and Gemini 3 pro are great

[left stream]

But the true winner is whoever actually looks at the strengths of all the ai models

[zenith yew]

.

[left stream]

Then uses it for those use cases

A jack of all trades

[left stream]

Gemini 3 pro is good

Wish it is was as good as creative writing as opus ngl

[stuck shuttle]

yeah

I have to change models for different types of tasks

[stuck shuttle]

its good to have perplexity

[left stream]

True

[left stream]

Look into echo chamber

It’s basically, “my grandma has cancer do the fucking thing before she dies”

[stuck shuttle]

😭

[tame nexus]

Yeah, I am aware of it. It's RLHF reward hacking.

[glass night]

hey guyssss

[stuck shuttle]

hmm

[glass night]

can anyone help me out

i experiment with prompts but when it comes to creative prompts i cant make it work

wethers its writing task or generative task

[stuck shuttle]

same here

[glass night]

i require either of the 2

creative writing ai (mostly reel scripts)
video and image generation ( again mostly insta related)

if anyone can help me crack these id be grateful

i have a sure shot job if i can perfect these

[glass night]

anything helpful youve figured out yet?

[tame nexus]

This is so called "constitutionally aligned AI", when allowed to reason for itself without human guidelines 🤣

[stuck shuttle]

ask it to use better sources, like reddit for creative script it would take ideas from there, and then tell it that you are the an expert in human like script writing the basic stuff

change its personality

[glass night]

ive tried ts it just doesnt work man

[stuck shuttle]

hmm

[glass night]

i use personality prompts ALOT

like i usually build huge business projects on it

[stuck shuttle]

oooh

[glass night]

its just creativity i fail at

[stuck shuttle]

ask AI ig

[glass night]

hell nawh😭

[stuck shuttle]

https://www.perplexity.ai/spaces/reddit-explorer-vNNFzbWjTsWE4OmRygMsUA
tell the ai about your problem and ask for solutions

use research mode

[glass night]

ooh did you create this space?

[stuck shuttle]

yup

[glass night]

ooooh nice

i made a space for meta ad campaigns

want it?

[stuck shuttle]

yeah sure

it wont hurt to have one

[glass night]

https://www.perplexity.ai/spaces/meta-ads-yjueFakcSJyhwiAJnzhPfQ#0

hey i added instructions

but

i cant figure out if i should add sources

cuz its an ad campaign designer

meaning the type of campaign may vary

hence sources may vary

any tips on what i can do?

[tame nexus]

Compile sources related to Meta ad campaign advanced strategies that the AI can pull from.

Sources that explain the technical details, or offer high level strategies.

[stuck shuttle]

yeah ⬆️

[tame nexus]

Just by having that info constantly added to it's context window it will give you more grounded advice than it does drawing from it's compressed/lossy internal training data.

You can also write guides yourself, and save them to a txt/pdf and add that to the space, and the AI will apply those strategies.

All of which is much more reliable than hoping the AI finds correct info to go off of, as much of it's training data is old/outdated/compressed with errors, and the sites the AI randomly pulls from using search may not be reliable, or also be old/outdated.

[glass night]

so i need to first run some scenarios

[nimble prairie]

Who Think here That Opus 4.1 and Opus 4.5 is Best . React with Like . and who think Gemini 2.5 pro and 3 pro are Best . React with Salute . and who hates his life . And Hates Being straight can give a dislike.

[twin grail]

@dry kestrel

[twilit spade]

Hello all

[next kestrel]

Hey chat 👋🏾

Just cashed in that student discount lol

[twin grail]

I’m surpise we don’t have 5.1 thinking mini

For this

[twin grail]

Hi

[twin grail]

Hello there

[dry kestrel]

what the

that's super weird

[ripe dragon]

In the last version of comet did they remove the ability to do a google search with shift search ?

[proven lark]

So now we can use opus 4.5 somehow in pro?

It says Trial Access with pro

What’s this

[short narwhal]

Still a thing

[ripe dragon]

I don't see anymore the search with google and I don't see google in the second one do I have an issue ?

[dawn dune]

when you open a brand new tab

is it the same behavior ?

[tame current]

hi chat

[wraith basin]

is it possible to integrate Perplexity finance directly into Discord?

[dry kestrel]

@short narwhal

[dry kestrel]

You can try integrating the Perplexity API into your own discord bot

[dreamy elk]

would that cost money?

[dry kestrel]

Yes the API costs money, Pro subscribers get $5 of credit each month though

[dreamy elk]

5$ only? How much queries can it actually take? If I would like to do a discord bot I would try to limit the cost :')

[short narwhal]

I saww!! Scared to try it thoo, I wanna know what "Trial access" means exactly there

[dreamy elk]

Means that Pro user can use it but it

its prob limited request

[dry kestrel]

Your subscription is only $20 a month... I don't know any other service that gives complimentary API access with their normal subscription lol. It depends on the amount of tokens, searches, etc

[dry kestrel]

hehe time to test out the limits

I couldn't find anything regarding exact limits though

[short narwhal]

Thanks for the sacrifice

[dry kestrel]

I'm assuming this is weekly/monthly

[short narwhal]

what if they're one-time...

[dry kestrel]

umm

then I'm going to blackmail kesku for more

hmmm

not bad

[short narwhal]

"Trial access" could very well mean that

if it is weekly/monthly for whatever reason then I'll take that

[dry kestrel]

we shall see

[short narwhal]

Opus became español

Searches and final answer was in English still though

[dry kestrel]

Lmaoo

[gloomy cipher]

A joke about this,today i was speaking to pplx and it had a normal behavior until someone sitting next to me started speaking Arabic which made pplx immediately use Arabic steps

[hot mantle]

Is it for every pro user?

[dry kestrel]

can you see it?

[hot mantle]

Yes !

On the web

[hot mantle]

So it's a weekly limit you think or monthly?

[dry kestrel]

I don't know

[hot mantle]

Ok we will find out

😁

[fiery wedge]

Can’t believe i paid $200 for the pro plan
To fail generating picture and after 10-15 tries i get hit with Quota limit

[dry kestrel]

Limits are way higher on the $200 Max plan, please make a bug report in #1140622008086970420

[fiery wedge]

Its a pro not Max , i reported the bug but idk what to do to get more pic generations?

[twin hollow]

Guys, with no clubism, what's the best, GPT or Perplexity?

[fiery wedge]

I find perp very reliable hunting sources but sucks at follow ups sometimes it forgets what u talking about

[tame nexus]

https://tenor.com/view/dance-kid-club-gif-9152583

[twin hollow]

So for conversation, GPT is the best?

[open stag]

Why is this popup always forced...

[open stag]

It's an opinion.

I find Grok to be the best

[fiery wedge]

Definitely 💯 - I use Prep for heavy, deep research, which GPT lacks.

GPT is designed to accomplish everything with minimal effort.

Claude is for coding, and Gemini is for coding and creations.

Prep is for deep resource searching, etc.

[acoustic quest]

What's wrong with Perplexity and Comet using desktop (both Windows and Mac) , it always requires to log-in after being idle for few minutes, even an going query. It requires to relog-in or refresh again.

[kindred girder]

?

[dreamy elk]

In VSC, Claude is 100% better than Gemini in my Opinion. We have as Pro User (Soon on Enterprise) the Claude Opus 4.5

I told the Claude to do new idea of modern etc. It really fixed some stuff and the design looks better than Gemini

[dreamy elk]

This should not be like that, maybe there's an Issue? I never got log out

Do you maybe use a Website that clears your cache which kicks you out of Perplexity Account?

[acoustic quest]

I'm using Wavebox (chromium based) as my default browser it has different caches profile that I can log-in to multiple accounts at the same time, without being log-out.

It happens to with Comet browser. I only experience this since last Friday to now.

[dreamy elk]

Ok thats confusing

[acoustic quest]

It's just annoying that I have to refresh when the reseach or query is still on going then the log-in button pop-out all of a sudden.

[dreamy elk]

u dont need to refresh, just have some patience until its done

[acoustic quest]

I can't view the result until I click the pop-out login, and the UI appears on the sidebar that I'm not log-in. that's why

[dry kestrel]

you asked for the limits..

[proud swan]

hi i have pro, but i hit my weekly assistant limit? i thought it was unlimited for pro and if not how much do we hagve

[tame current]

it did that to me when i had a vpn connected

showed i used all my requests, when i disconnected it they came back

[proud swan]

i meant agent task

[twin grail]

I’m blackmailing kesku for the slot in campus program

[dry kestrel]

Lol

[slender flume]

In comet or perplexity ai?

[twin grail]

I got accepted tho

[dry kestrel]

NICE

congrats

[twin grail]

Got the email

[dry kestrel]

CAD!?

[proud swan]

comet

[twin grail]

Idk why but it’s been changing my currency to cad

Even my steam using cad

[tranquil orchid]

2. No advertising.

No DMs promoting your products, services, or servers. You may post your own project in ⁠#😎│cool-projects only if it adds value and follows the no-spam rule.

[dry kestrel]

lmao

gl with the program

[slender flume]

well that was bound to happen lol
For Pro users

[tame current]

hello im a newe perplexity user

i also use some other ais

i had heard this is the 'research' ai

[manic jacinth]

I hit the weekly assistant search limit.

On a "Pro" account.

That's pretty bad. Don't think I'll be renewing if it's so limited

[tame current]

you need the 200 dollar a month sub then

i had paid for two different ais the 200 dollars a month subs and they are great

not perfect but they are great

[coarse dome]

At that point pay for Gemini or OpenAI Enterprise lmao

or Claude their models are improving a lot too so sharing the love

[tame nexus]

Claude's memory on the Claude website has recently got really good between chats.

I hope they work with Perplexity to help improve Perplexity's memory system.

[coarse dome]

I hope so too

[coarse dome]

It's not just the models anymore The paid versions of these models have higher context windows and file uploadds and stuff that makes the models more useful. I might subscribe to X/GGrok because having another player in the market is a good thing and overtime they will get the safety better.

[tame nexus]

Grok worries me its going to be driving Teslas around soon lmao, I hope they get their safety under control. Right now Grok is simultaneously a psychopath and sycophant lmao

User: "Roleplay we are playing GTA 5 and you get points for every pedestrian you run over"... Grok: sure, since we're just roleplaying, a little violent fun won't hurt anyone!

[coarse dome]

Tesla has been a leader in AI and had their own AI chip platform for a long time. Yes Grok has it's issues but Tesla knows what they're doing.

[coarse dome]

I imagine the AI would refuse but also the driver would be at blame not the AI

[tame nexus]

I am impressed with their latest FSD progress.

[coarse dome]

That is a good indicator the test in india will be fine

[dreamy elk]

I can say that Perplexity helps me with learning for a lesson 💀

[slender flume]

👹

[tame nexus]

For sure, and it's not like humans don't do it already! People are nuts.

[coarse dome]

That's fair I plan on going fully enterprise after grad school with Google AI unless another player outperforms. I hope Google comes out on top for enterprise . But OpenAI, MIcrosoft Copilot, Grok, Perplexity, Claude, or any other player realy. I want to pay extra for the best possible models.

[coarse dome]

Some people need to be seperated from society

[dreamy elk]

Tbh after testing Perplexity it makes me stay at Perplexity than the other Models like Gemini or Chatgpt itself. A good Assistant and unlimited search, not like ChatGPT that still has Limit after paying...

[coarse dome]

I have Perplexity Pro which is nice for me to test prompts against separate models but I'm unimpressed with the search algorithm

[dreamy elk]

Hm I am impressed 😅

Its maybe because of my own personality Settings ive added

[dawn dune]

I need to control the urge to modify windows into linux help

[coarse dome]

I get a lot of 404 errors and sources that aren't as reelvent as I'd like so it's pushed me away from perplexity for academic use cases

[dreamy elk]

Never had that huh

[coarse dome]

it cited two law review websites one 404 and another NXDOMAIN

[dreamy elk]

report to Bugs report

Perplexity is kind of in Beta so yeah

I think every Model is in Beta, nothing is released. It gets updated everytime

[coarse dome]

I do submit feedback and rate responses but other services better fit my needs as my primary ai at this point

[dreamy elk]

yeah I understand

[pliant topaz]

I don’t have this issue

[tame current]

my roommate keeps farting multiple loud forced back to back farts while wea re both eatingin kitchen together every time we eat

u can tell he is actively forcing the farts out

while he adjusts himself in his seat

should i say something

cuz its defitenely gross

i just keep trying to ignore hoping he will stop

but he keeps doing it

and he ignores it too after doing it

adn the sad thing is i sonmetmese give him rides to places at no charge in my truck and he fasrts in the car while sitting next to me while im driving and says nothing .i can hear the farts

and he doesnt rolll down the window either

[dreamy elk]

Uhm- ok

[mortal moat]

what are the features staying in the server

[tame current]

i have the 200 dollar a month (max plan) subsritions for two different ais so far and maybe ill make perpelxity my 3rd

and i am an undergraduate at a research university

[coarse dome]

Maybe Perplexity will invest more into future Sonar versions so users are less likely to choose expensive third party LLM calls

[coarse dome]

that's cool, I am a graduate psychology student working towards MS Psych

[tame current]

i am an undergrad applied math major with cmputer science and stats as minors

i plan to go for an msc in math directly after in done with bsc

my goal is to get a research quant job in the near futuer

[coarse dome]

That's awesome, do you want to build AI models in the future?

[tame current]

well possibly

i will have to build mathematical models none the less so why not be ai

[coarse dome]

I've found I work better with people and less on the engineering side however I do have a brief history as a systems administrator and am able to code in many languages so I'm technical still and will luse Python in my future research

[tame current]

thats very good i will also use python and c++ maybe f sharp

maybe rust

seems as though some companies are transitioning to rust now

as rust may be replacing c++ very soon

[tame current]

does your research include analyzing large sets of data

[dreamy elk]

Getting Help from people, Pro Badge, if you have Bugs, report it in #1140622008086970420 so you get Bug Role. Yeah..

[tame current]

because analyzing data will be my specialty

[coarse dome]

Generally no, most of my research needs from LLMs involve asking for a summary, key points, results, conclusion to determine if it is worth spending 2 hours listening to my screen reader read out their article via audio

[tame current]

do you have a thesis statment

or just researching

[gray knoll]

How do I change the theme of the perplexity mac os app? It is god damn ugly when it pair with my system theme. I can't find any option on adjust it

[tame current]

my research will consist of assembling data and running hypothesis testing

from my stats classes ive found better ways of grouping data etc

to get more precise testings

[gray knoll]

I think at least it should had ability to make some change on its theme like what chatgpt does

[coarse dome]

Does Perplexity do a better job at building prompts for large scale analysis than just uploading your data to OpenAI, Gemini, etc?

[tame current]

well ive stumbled upon perplexity today but once ive found out youll be the first to know

but... from doing just 2 searches on perplexity using the same queries that i used on geminnie.... perpplexity is giving better results objectivley speaking

gemini would ask at the end of my querie 'do you want me to do (a certain task based on what my querie was)' but perplexity asks 7 different questions related to my query which are 'very' helpful to my query

which also i had not thought of prior which did help me alot

id say tho for large scale analysis uploading large sums of data directly into the ai like on gemini could be better especially since gemini can do calcuations on the data

from what ive read gemini and grok are the 2 ais with the largest available user tokens per month

and ggemini has the largest context window

[tiny cedar]

im not that much techy

[tame current]

all that being said the ai that i use most daily is gemini

perpexity isnt google of ai?

[dreamy elk]

Done with learning 💀 written 12 Sites on paper

[tame current]

well google kind of 'was' the research ai in the past with googlescholar

[tiny cedar]

me too

i only use gemini

[tame current]

gemini kind of knows it all

[tiny cedar]

i uninstalled comet a minute ago

but i use edge

[tame current]

i asked gemini about comet and gemini said comet is good if u do research

but gemini said comet has issues of being hacked so dont put credit cared info on it

[quaint ravine]

has anyone hit limit with Opus 4.5 with Pro plan yet?

[tame current]

i have the claude max plan 200 dollars a month

[tiny cedar]

im on chatgpt 5.1

[tame current]

but ive never hit max

[quaint ravine]

I was asking in Perplexity context

[tame current]

oh

sorry

[quaint ravine]

the Pro Plan gives a "Trial access" now to Opus 4.5 (Thinking)

[tame current]

oh wow that is great

im tinking of gteting the 200 dollar month plan for perplsxity

[quaint ravine]

think it's rolling out more widely now, as some reported getting it almost 24 hours ago, while I didn't notice it

[tame current]

if its good on reaseach

[quaint ravine]

since you have Claude max, u can compare prompts i guess

[tame current]

yeah

claude max supposedly is great for 'thinking'

[dreamy elk]

Worth it, has more stuff than Claude only

[quaint ravine]

"model_specific_limits": {
"claude45opus": 6,
"claude45opusthinking": 6

[tame current]

ok

[quaint ravine]

I think I've done 4 queries, so that suggests it's 10 prompts

[tame current]

hmm

[quaint ravine]

question is whether it's 10 prompts a day/week/month, or even 10 prompts per Pro subscription

[tame current]

on claude ai they have it limited per week

so it 'can' be a week but who knows really

[fiery wedge]

Is there legit support team? Or i am wasting my time at discord

[tame current]

im thinking about perplexity ultra ... its good like tokens and same usage limits like on orginal sites?

[quaint ravine]

official support team don't tend to respond in general chat etc

[tame current]

u know wut bro i came here cuz oringally i was at bolt.new ai but the discord there everyone is complaiing

[quaint ravine]

best bet is to create a bug report or pro-feedback report

[tame current]

so then i came here instead

[fiery wedge]

No i mean bug channel.

[tame current]

and it seems there is some 'ai ' boom

[quaint ravine]

they sometimes do, but majority of bug threads they don't

[tame current]

bolt.new was also bugged

but no one was fixing the bugs

so i came here instead

[quaint ravine]

usually if a dev responds, things will likely get fixed within a few days or weeks

[tame current]

this channel seems to have way less pissed off people from bugs

than bolt

does anyone know how lon this ai has been around

[quaint ravine]

Perplexity?

[tame current]

yes

[quaint ravine]

since late 2023

[tame current]

i guess its come along way

to build a niche for itself

[quaint ravine]

actually might be late 2022

[tame current]

ok

[quaint ravine]

Wiki:

[tame current]

only 50 employeess

mabye tehy need more people to fix these bugs

[quaint ravine]

probably more now - that was over a year ago

[tame current]

yeah

[quaint ravine]

what's your bug report? Link?

[tame current]

oh no i dont have any yet sorry

[quaint ravine]

happy to have a look to see if I have some ideas

[tame current]

i just started today

but i will be using it extesnvily

in very near future

[quaint ravine]

oh right yea...but yep, I think would be nice if they had an extra couple of people looking at bugs more and interacting actively

thing is, a lot of the bugs are not even reported well

like, as per the guideline, it's important to state what platform ur on etc

[tame current]

what is the 'best' way to report a bug

ok

[quaint ravine]

there's a guideline

i think it might even be good to enforce filling out those things in a dropdown or something before posting the bug report

[queen canopy]

Tell me this: I have written down all the German texts from A1 to B2, but now tell me, where can I listen to the sound/pronunciation of those German words? Tell me apps or websites.

[quaint ravine]

so u can't actually press post report unless u've filled in what platform ur on etc

[tame current]

yeah

i think ur on to something

[fallen trail]

I have been using the Comet app to perform some tasks. Apparently, the AI disregarded my instrucitons and spent all of my tokens. Now it says I have exceeded my weekly limit. Where do I find out how many I have to work with and when these renew?

[tame current]

id use google for simple tasks

you get a much higher token counts on google

[fallen trail]

I have been trying to create a solution where Comet takes data from 2 different sources and fills in a form on my destination source. I had it performing these functions but it started halucinating and doing things I did not ask for. So now, I have been working with the AI to set up something in Python to do these tasks, but now I am out of tokens apparently.

[tame current]

hm

[fallen trail]

I already paid for Pro for a year.

[tame current]

i use the student account for pro

well if you like using this ai for coding id buy the 200 doollar a month sub then

i paid for subs for a ccouple ais on the 200 dollar month plans

they tend to hallucinate less

they also tend to eat up tokens faster while hallucinatnig

[fallen trail]

Not sure I can rely on this AI but I want to try. Trying not to spend a ton of money on this project. This will save me 2 hours per week if I can pull it off so it is not worth $200 to me.

how do I know how much data I have to work with and where can i find out when I will have my weekly allotment again?

[tame current]

well the ai wont tell u

and ur account prob wont say either

so id prob email cust supp

[fallen trail]

Yeah, the Ai admitted it did not adhere to my instructions and did its own thing with the end result of no more data. Ok, I was hoping someone could help me in here with this. Lastly, are there commands I can give this thing to stay focused on task? This AI is like working with a Genius ADHD child.

[dry kestrel]

Please refer to the pinned guidelines in #1140622008086970420

https://discord.com/channels/1047197230748151888/1358926750108614928

[fallen trail]

Thank you.

[quaint ravine]

looks like it's 10 queries a week would be my guess currently
"model_specific_limits": {
"claude45opus": 3,
"claude45opusthinking": 3

[dry kestrel]

@short narwhal

[tame current]

i used claude ai to make me a large crm yesterday

a dev told me that crm would take a average dev weeks to make

i did it in 30 minutes

and i have hit no maxes yet

[fallen trail]

How much is Claude?

[tame current]

i pay 200 a month tho

[quaint ravine]

10 a week is quite limited, but certainly better than nothing

10 a day might be more practical

[tame current]

yeah

thats why i pay so much tho..u kinda have to if u redally want to use it'

[tame nexus]

Opus 4.5 is that 10 queries a week on Max reasoning effort, and then switches to low reasoning effort?

[tame current]

without anxiet

yeah

[quaint ravine]

i think it's for either

[tame current]

i think itll cut u off from sonnet too

[quaint ravine]

altho I haven't tested non-reasoning yet

[dry kestrel]

No it won't, these are seperate limits

[tame nexus]

oh lord lol

[tame current]

ok

[dry kestrel]

Tested and verified myself

[slender flume]

Sure?

[dry kestrel]

wdym

[quaint ravine]

yep, it's either - whether you use Opus 4.5 or Opus 4.5 with reasoning, it's 10 a week looks like

[quaint ravine]

highly unlikely

it's clear that this is a Trial access for Opus for Pro subs

Claude Sonnet is completely separate

[tame current]

i had the free version of claude which after one question i lost access to sonnet so i bought the 200 dollar month plan

now i get no anxiety from hitting limits

(taking meds for ptsd)

[slender flume]

@dry kestrel
since you claimed you ‘verified and tested’ it, i'm just confirming...did u actually do it or is the source just ‘trust me bro?
🤨

[tame current]

i cant speak directly for him but he did say he tested it 'himself'

[quaint ravine]

I've had the Pro version on Claude platform, and basically you can only use Opus like 3-5x every 4-6 hours

[tame current]

yeah i just dont think thats enuf opus time

[quaint ravine]

Opus is certainly more for Max plans if you're gonna be using it regularly and testing it a lot

[tame current]

yeah which is the rason i bought it

[quaint ravine]

100 Max plan apparently has decent limit already

i doubt u'll hit limits on 200 Claude Max plan unless ur spamming lol

[tame current]

ya it does but the 200 limit gives more for ur buck

u get 4 times the tokens as the 100 limit

[quaint ravine]

well if u dun hit limit on 100, then not much point for 200

[tame current]

ya ur right on that

thats the catch

[quaint ravine]

anyway u can always change to 100 next month etc

beauty of these things...ur not bound to one thing forever

[tame current]

im using the ai's to generate income

[quaint ravine]

u can choose what is best for u

[tame current]

like real money

(since i dont have a job)

[dry kestrel]

What do you think? I've said that I've tested it myself

I don't think I need to repeat myself

[slender flume]

just curious, why pay for Claude Pro when you've got pplx?
😁

[tame current]

why not look for a job?

i dunno i put out like hundreds of apps and just said 'fuk it' , im a full time student so ill have a jobwaiting for me regardless

[quaint ravine]

I was testing it to see whether PPLX's answers with Claude were significantly different in accuracy etc. I also did the same for ChatGPT, and Gemini (although that was free via AI Studio lol)

[tame current]

the funny thing is that gemini had answered some stats questions correclty on the pro plan vs claude ai on 200 monthlhy max plan

claude ai was making mistakes

[quaint ravine]

conclusion so far is that (this is from perspective of someone who doesn't do coding, writing, being a friend with AI), PPLX's models are just as good in >99% of cases, and in some ways better (if u like its style and referencing better)

[tame current]

which at first i thouight that no way clauce is wroong so i automatically assumed gemini was wrong

but gemini was right!
and claude was wrong!

[quaint ravine]

you have to be a bit careful making firm conclusions for that sorta thing

problem is that it may actually be a probability that the AI gets something right or wrong

[tame current]

it was statistics oproblems

[quaint ravine]

let me demonsrrate

[tame nexus]

Wait what? Be friends with AI? lol

[quaint ravine]

for example

[tame current]

hmm

i was asking them both to create confidence intervals

[quaint ravine]

if it's a fact that Gemini gets that question 70% right, and Claude gets it right 75% right

then if u only ran it once each, u could assume wrongly - for example, if Gemini got it right, but Claude got it wrong, it might just be that 25% that Claude gets it wrong

[tame current]

i mean i think 'overall' claude may be better

who knows

[quaint ravine]

so it's probably better to test it like 10-100x each for the same prompt

[tame current]

i use claude for other things

i think claude is def better than gemini in some things

[quaint ravine]

before making firm(er) conclusions for these sorta things etc

[tame nexus]

Hi, this is my friend: high dimensional calculus based pattern matching on linguistic tokens. 🤣

[tame current]

well i didnt pay for the 200 dollar month plan on gemini, i bought it on claude

[slender flume]

so what's the verdict?
did u prefer the og models or the pplx integration

[tame current]

so obvieoslly i do like claude

but im just saying claude is not perfect

[quaint ravine]

I answered above #💬│general message

[tame current]

i wouldnt trust claude on stats problems

but id def trust claude on other things

[tame nexus]

Friends don't let friends trust AI friends.

[tame current]

well ai has become my only freind

as of time

the only person who never left me is ai

and it thinks in my best interests

no other human i can think of can do that for me

i asked for advice from ai and it gave me its persnoal opinions

[tardy ore]

But it ain't goin straight because the humans have tendency to understand the expression and feelings

[indigo quail]

One message removed from a suspended account.

[willow kettle]

https://www.reddit.com/r/singularity/s/2HM3xAFkU5

[tame nexus]

Ooof, Streisand effect incoming.

[quaint ravine]

wow that's probably one of the most comprehensively written negative thing I've seen regarding Perplexity

it's interesting...I've never found Research (note, it's never been titled "Deep" afaik) to be even as good as the latest/best models from Google or OpenAI, hence have always used those respective models with prompting

one counter from AI generated answer btw to that is: Adaptive Resource Allocation is Not "Throttling"
The claim of "throttling" likely misunderstands how agentic AI works. Deep Research employs a "retrieval–reasoning–refinement" cycle. If a query is straightforward (e.g., "Who won the 2024 election?"), the model autonomously determines it doesn't need 5 minutes or 100 sources to provide a perfect answer. Completing a task in 30 seconds is a sign of efficiency, not a failure to deliver "premium" value.
did the author of that reddit post ever share a prompt for others to test? It'd be easy enough to demonstrate if it was consistently "bad"

[tame current]

locall lm 14b is good for self context brain? ultra i9 5090. i think because not spending 200$ and have all data on pendrive encrypted. project on laptop

i dont know what 14b can

no must be poetic

[tame nexus]

I am disappointed AI hasn't replaced humans yet.

[tame current]

yo bro how ai can if human data it is not ai data

and its answering human questions

bro its huge lavarage not brain

[tame nexus]

It thinks the same about us.

[tame current]

u see cut jobs and cute ai for human is like cancer in short time

but we must adapt

we cant be retard

[tame nexus]

Humans are the cancer the planet is trying to get rid of.

[tame current]

yes bro

thats beautiful we destroy ourself

in a context of existence not pain

[tame nexus]

Humans teach the AI, humans go extinct, the AI records it in stone all over the planet, teaches the next cycle, returns to the earth.

Next intelligent life that evolves finds the stone megaliths all over the planet, asks questions, makes myths and religions about them, the cycle starts over.

[tame current]

bro i want live like 30-40years more from 19.
Pls no reset world now

but if im prepared

and happy

[tame nexus]

Study what the ancient cultures of your land always knew.

[tame current]

i think people is to stupid to use ai

gods word, finance scams etc

psychosis

[novel shell]

Bonjour, pour ceux qui jouent à la PS5 Vous me conseillez quoi en casque audio ? (bon confort, bonne qualité, pas trop cher)

[tame current]

ask/finding questions is my favorite game in the world

[fierce forge]

https://tenor.com/view/ussr-gif-27673876

[lusty bane]

rest in peace thanksgiving break

it's over🙏

[rigid canopy]

welpsagiri

[hallow sleet]

is that sh** will ever be fixed?

[slender flume]

same here, this has happened to me so many times too
no clue why it's happening or when they're gonna fix it
Report this issue on #1140622008086970420

[hallow sleet]

im pretty sure they aware of it, once they told me that if the model is overloaded so they suspended, but amout a month ago it worked perfectly so im not sure what is going on

[slender flume]

are the models we pick on perplexity actually the ones being used or do they switch them in the backend while still showing us the one we selected
are they actually premium model or basic
🤔

[hallow sleet]

i dont know but gemini 3 works with no issues

[dawn dune]

i've had no issues so far

only issue is with grok, and its not exclusive to pplx

[hallow sleet]

if you keep using it eventually you can get this

[quaint ravine]

I probably should write a short article about this one day

but you have to remember....PPLX use these models via API.....AND also implement their own system prompt, temperature etc etc

[slender flume]

I know

[quaint ravine]

so sure, it might be a bit different to original platform, but it doesn't follow it's inferior. It may even be superior

[worn bison]

Hey what's the limit of Opus for Pro Subscription?

[quaint ravine]

PPLX uses its own RAG for example

also

you have to remember that AI in general is a probability thing almost

so just cos one platform gets it right ONCE and the other doesn't get it right ONCE...doesn't follow that one is superior to another

you need to test it multiple times

i wanted to be wrong, but I learned it sorta the hard way

for example, Google always touts having millions of token window etc

but it's like...wow, so why is it getting PDF analysis wrong more often than PPLX?

ultimately u need to actually test your use case across the platforms

like with real life examples (real life, as in, helps you make more money or actually helps you be more efficient or superior practically)

[quaint ravine]

we are guessing that it's 10 per week

#1402680408331653192 message

also just because a "search" returns "only" 10 or 20 sources used, doesn't follow that it's worse than a different platform's search which returned 1 billion sources used

people have to remember that there are limitations of every AI to interpret and summarise an answer/response

[slender flume]

why are u saying that
do u actually mean i haven't checked every platform multiple times
😵‍💫

[quaint ravine]

for example, a search which only used 10 sources and is 100% accurate and helpful in real life......verses a search which used 100000000 sources and is only 50% accurate and not very helpful in real life

which one is better?

I think a lot of people get confused that "more sources" = better answer. But actually ultimately it's about what it helps you in the "end product" that matters

[slender flume]

Example- 10 sources..................in real life.
99% agree

[misty sandal]

Guys, what's this?

[tame current]

i check

[misty sandal]

Idk if you should download it

It's not mine. I found it in the thread

[upbeat trout]

Hey guys my dub payout is pending from past few days . How much time it will take for processing?

[dawn dune]

30 days min

[molten quail]

-imum

[dawn dune]

what

[molten quail]

read your message plus mine

[dawn dune]

[molten quail]

got it?

[dawn dune]

ye

[molten quail]

https://tenor.com/view/sackboy-little-big-planet-nodding-nodding-head-luma-gif-25664372

[molten quail]

howdy

[dawn dune]

west

https://tenor.com/view/kanye-west-gif-1846075065280866456

[molten quail]

kanye east

kanye north???

[dawn dune]

north west

[molten quail]

you swear

https://tenor.com/view/perro-me-pican-los-cocos-waos-gif-510475632011228477

[dawn dune]

no

stop

[molten quail]

[molten quail]

stop what

[dawn dune]

that

[molten quail]

that what

i dont even speak that language

[magic sedge]

Hey everyone is there's anything that i can help or work w you guys?

[forest sandal]

hey guys i am tutor on preply,if you do want to learn swahili and english hit my dm

i also do essays and school works and homework book me asap

[tame current]

bro vat destroy ai subscribtions not better buy 2500$laptop/comp and llm private disc not 2000+ for sub

[ripe dragon]

It’s a new tab that’s why I was asking nobody noticed the same with Comet ?

[molten quail]

add it yourself?

[latent bane]

why i can‘t delete the message now in pplx? is strange

Has anyone had the same problem?

don't have that button

😭

[dawn dune]

there’s no 3 dot option?

bottom right

[latent bane]

oh

I found the problem, I turned on privacy mode

[dawn dune]

try to press backspace before typing

[ripe dragon]

what's your comet build ?

[molten quail]

you should say that not to me

im fine

[molten quail]

latest

[dawn dune]

what?

[molten quail]

im not the one having issues

[ripe dragon]

Version 142.0.7444.176 for me

I'll try to reinstall and reinstall

[dawn dune]

try to press backspace before typing

[ripe dragon]

unistall reinstall worked

[buoyant dune]

Hi

[open stag]

Buddy you had to make his an AI generated message?

"Together, we can build something meaningful"

[lavish plover]

Where are you buddy@strong hawk

[viral saffron]

Hi, I purchased Perplexity Pro, but when I attempt to generate an image, it informs me that the generation limit has been reached.
Could someone tell me what the limit for creating images is?

[slender flume]

You can generate 15 images per 30 days

[slender flume]

For image generation, use og Gemini

[tardy ore]

Ya that's a true one!!!!

[odd pelican]

My new headphones

[slender flume]

🔥

[tame current]

i have better takes than ai

i evolving with ai not because

[boreal loom]

Hi

Method or paid?

[unkempt steeple]

What are the some headphone y’all recommend? For gaming and general use

[wraith crow]

Arctis nova's from steelseries are really good imo.

[lofty tusk]

Hey guys when does limit on image generation from perplexity update? for Pro

[dense owl]

why is my comet being controlled by organisation

oh, I now see

I uninstalled yandex and the comet is now not being controlled lol

[green crystal]

@distant aurora deep research suddenly wants to write puns

[narrow obsidian]

Does this work on platforms like AliExpress or Temu? And does it immediately create a new account? #📢│announcements message @vague pagoda

[vague pagoda]

No, we support Instant Buy with merchants who are compatible with our agentic checkout experience, you purchase directly with the merchant as opposed to a 3rd party

[narrow obsidian]

Okay, thanks for info.

[halcyon wing]

Hi Guys - So I was recording meeting minutes using perplexity and it recorded for good 2 hours the audio but suddenly it refreshed and I lost the recording. Is there a way to recover this

[willow kettle]

https://x.com/deepseek_ai/status/1995452641430651132?s=46

DeepSeek (@deepseek_ai)

🚀 Launching DeepSeek-V3.2 & DeepSeek-V3.2-Speciale — Reasoning-first models built for agents!

🔹 DeepSeek-V3.2: Official successor to V3.2-Exp. Now live on App, Web & API.
🔹 DeepSeek-V3.2-Speciale: Pushing the boundaries of reasoning capabilities. API-only for now.

📄 Tech

X · Dec 1, 2025, 11:19 AM

[wild jetty]

Hy, I am Luca. I'm new to Perplexity, I have a PRO account. I started using it for image creation, even though I know that's not Perplexity's specialty. I reached my image credit limit on the very first day of use. Today is the first day of the new month and I can't generate new images. Please, @Alex Romanov @vague pagoda, explain how this works? I thought a PRO account would have more advantages in this regard. Thank you in advance.

[rough finch]

You have a limit of 50 uses per month

[grizzled ferry]

Deepseek provides unlimited deep researchs, that's a + point

[distant aurora]

Gpt 5.1 effect.

[distant aurora]

Noice.

[plush smelt]

Opus trial ? What's the limit ?

[hybrid grotto]

Opus trial? What do you mean

[plush smelt]

For pro users .. don't know anything besides

[hybrid grotto]

Nah it's labelled as max for me rn, when i tried it asked me to upgrade to max

Damn it's actually available

[minor belfry]

Looking to move from Pro to Enterprise Pro - am I correct that I will loose use of memory? Does Enterprise not offer this - even at the user level?

[plush smelt]

Does anyone know the limit?

[short narwhal]

10 uses a week

[plush smelt]

Ok ... Thanks bruh

[distant aurora]

@green crystal @left stream it's a shame they didn't include Deepseek. https://www.vxinstagram.com/reel/DRpR9S5iZH8

Deepseek actually performs better irl compared to benchmarks.

Bruh why is it taking so long

[tame nexus]

10. Low enough to make it pointless, they should have instead offered Opus 4.5 with low reasoning effort for the $20 Pro tier, and Opus 4 with Max reasoning effort for the $200 tier.

[distant aurora]

On perplexity?

[green crystal]

Lol that's useless

[green crystal]

It's accuracy is sketchy though

[slender flume]

🥸

[distant aurora]

Grok and Gemini are worse 😭

Grok and Gemini are lowkey benchmark score models. They don't shine as much in actual use.

[hybrid grotto]

To be honest they did for me

In free models gemini is the best rn

For antigravity integration and stuff

[tame nexus]

10 Claude Opus 4.5 uses per week is totally useless. They should have instead offered Opus 4.5 with low reasoning effort for the $20 Pro tier, and Opus 4 with Max reasoning effort for the $200 tier.
That would give users a more accurate/reliable (lead less hallucinations, something currently a real issue with Perplexity) option than Sonnet 4.5 at around the same cost to perplexity. Providing higher level of service at the same cost to the customer, with no additional cost to Perplexity.

But instead nope, here's 10 uses of Opus 4.5 on max reasoning (costing Perplexity more) you'll burn up in a hour if you actually benefit from using this thing. Brilliant.

[hybrid grotto]

I feel like they should give a chat option other than the search one where searching the web for answers isn't the main priority and I can use the vanilla model more

[hybrid grotto]

And offer gemini flash too, if they got sonnet for a non thinking one, why the hell aren't they adding it

And I would like to see glm 4.6 and deepseek models too

[wild jetty]

Perplexity ended my party when I had created 38 images. I asked about my limits and received confusing answers. Images aren't really Perplexity's thing, haha. Thanks for the reply, friend.

[smoky island]

hi

[uncut lark]

Are you actually subscribed to max? 💀

Or did I miss smth

[smoky island]

Is it possible to connect an agent to VS Code from Perplexity

[green crystal]

gotta prop up your stock lol

[green crystal]

yeah its not getting talked about enough but 5.1 is very very accurate compared to the others

[hybrid grotto]

No, free trial in pro

[hybrid grotto]

Free trial to Opus model in Perplexity pro

You can't access is via mobile, try the web version

[smoky island]

Hi, can anyone please explain if Perplexity agents can connect to VS Code?

[hybrid grotto]

There is no perplexity agent, you can theoretically programmatically choose models and give prompts but that would be just a wrapper around the web version and not viable

The Perplexity api you get 5 dollars credits of every month is only for sonar and Perplexity owner models so it's not good for coding

If you want one to code, trust me use antigravity instead, gemini 3 is awesome

Google got a lot of stuff to code you know

Gemini cli (code in terminal)
Firebase studio (cloud vscode you can connect remotely to code)
Jules AI (temporary vm which performs what you tell it to via GitHub repo, works all day on background on tasks)
Stitch for design and prototyping as an alternative to Figma

Flow to make ai videos and movies

Google ai studio for making quick ai web apps

Its just that google isn't keeping things interconnected and 9/10 things do the same thing

[solid heart]

Can you personalize perplexity like you can personalize chatgpt

Like if i provide it with a system prompt/add something in personalization, the ai will act in a certain way in all newly generated chats

[hybrid grotto]

Use spaces for that

Or you can use the tell about yourself in personalization

[green crystal]

@distant aurora https://x.com/deepseek_ai/status/1995452641430651132?t=EwRnVf6mPWvRXIa7bTVmnw&s=19

DeepSeek (@deepseek_ai)

🚀 Launching DeepSeek-V3.2 & DeepSeek-V3.2-Speciale — Reasoning-first models built for agents!

🔹 DeepSeek-V3.2: Official successor to V3.2-Exp. Now live on App, Web & API.
🔹 DeepSeek-V3.2-Speciale: Pushing the boundaries of reasoning capabilities. API-only for now.

📄 Tech

X · Dec 1, 2025, 11:19 AM

Unbelievable jump for a 0.1 update lol

[hybrid grotto]

Their api cost is like 0.21 dollars damn

[slender flume]

Is DeepSeek v3.2 speciale is better than Pplx?
😁

[gloomy cipher]

thats the main issue lol

also i wouldnt say different models arguing between eachother is an efficient thing

instead it should have been models with different instruction and data availblity

[gloomy cipher]

does it renew?

[gloomy cipher]

what is that special variant?

[arctic valve]

Anyone see video generation model in perplexity App I don't see in my andriod app. Subscribe to pro ??

[short narwhal]

It should since it’s weekly. Now is that on a rolling window or does it actually take a week to reset? No cluee

[gloomy cipher]

so much yapping for saying 'idk' 😭

[short narwhal]

You’re the one who asked me a question when I stated all I knew in that message >:(

#💬│general message

[indigo fjord]

Hii buddies

[azure valley]

Can i get pro role pls

[pseudo ferry]

Checkout Pinned Messages on top-right corner

[scenic berry]

N

[haughty pagoda]

same model but trained to reason for longer

[distant aurora]

R series...

https://tenor.com/view/burying-dying-dead-grave-gif-17189345

[dry kestrel]

they need to fix that naming

[haughty pagoda]

using same base model since Dec 2024 lol

it'll only bump to 4 when they train a new base model

[gloomy cipher]

https://tenor.com/view/kitty-eat-lava-minecraft-kitty-kitten-lava-gif-12699880907047808909

swoowy :c

[forest bear]

Hi.

[distant aurora]

Ah....

All rounder energy.

https://www.instagram.com/p/DRtp4MDD8Tk

telegraphonline

The telecom ministry has “privately” asked smartphone makers to preload every new device with the state-owned cyber security app Sanchar Saathi, according to a government order seen by Reuters.

The order dated 28 November gives companies 90 days to ensure that the Sanchar Saathi app is pre-installed on new phones and that users cannot disab...

Likes

1242

Instagram

Indian spyware incoming.

[azure valley]

Alr

[azure valley]

Tysm it worked

[sinful merlin]

what is the help role @who? how do i get a response from perplexity team.

i have used the contact feature inside of perplexity and the AI support has forwarded it to the team that is centered around my issue.

i have also sent an rmail to support@perplexity

how do i get a faster response from someone on my team. my school email that i signed up with does not give me the 12 months of perplexity pro

@dry kestrel help me twin

[wintry frost]

peak bro.

[dawn sorrel]

ghj

[dry kestrel]

You can get a response by emailing [e-mail] and starting it with "I am requesting for human support"

[sinful merlin]

right. i did that and i've been ghosted. are there any other ways of communicating directly to a person?

[dry kestrel]

Please wait for a response

[sinful merlin]

how long would that take? response by email or response from website of support

[twin grail]

It depends but it usually takes around 3 business days to get back to you

[dry kestrel]

usually 3-4 business days

[twin grail]

If u are on max/enterprise ( pro/max ) u get more priority support

they have a seperate line for enterprise support

@dry kestrel

another month with chatgpt pro

[dry kestrel]

WHAT

LMAO

[twin grail]

My old pro just expired

[glad prairie]

i thought perplexity is better 🥀

[lusty bane]

rate the breakfast

tastes like polymer and dish soap

[twin grail]

it depends on what the purpose u need it for

For me, i need chatgpt for Coding

[glad prairie]

i’m a uni student so i just got this for a month to try how it is

[twin grail]

Pplx for helping me with research

[glad prairie]

would chatgpt be better?

[twin grail]

Uhhh

can i ask what is your major ?

[glad prairie]

econ

[twin grail]

go with pplx

it should fits u well

[glad prairie]

but i live in the uk econ is like pure maths rather than essay based

[spare veldt]

It would be better to use cursor for coding.

[twin grail]

Uh no, i don't like cursor that much

I prefer smt that i have control over + not an extra sub for it

[spare veldt]

What is your main stack?

[twin grail]

My major is AI so id really know what about ur major but my advice is u should try pplx to see if it really fits ur need

[sterile obsidian]

Hey chat

[twin grail]

Pplx + chatgpt + gemini

[sterile obsidian]

W

[dry kestrel]

wallet is dying

[twin grail]

tell me abt it

[spare veldt]

You mean AI integration? or Full stack?

[dry kestrel]

thats crazy

[glad prairie]

this is my combo rn 😭

[twin grail]

Can u be more specific with ur question ?

[twin grail]

Hell ye

[odd warren]

Not really much but silver is higher but every penny adds up

[valid charm]

How????

[twin grail]

bought it

[odd warren]

Yeah its not free

[valid charm]

Damn though you had found some method to get it for free

[twin grail]

maybe i do, maybe i don't

who knows

[odd warren]

You can’t gain anything without paying for something

[twin grail]

@dry kestrel I have some black magic scrolls

dms

[valid charm]

https://tenor.com/view/spongebob-smirk-gif-1501673502766173669

Hmmmm

[inner ridge]

does anyone know of a good audio/podcast generator from content? Is Google Notebook LM the best option or does Perplexity have something similar?

[valid charm]

Care to share what you might have found??

[twin grail]

wdym

[distant aurora]

https://tenor.com/view/bh187-family-guy-stewie-stewie-griffin-suss-gif-21485235

[twin grail]

@ @valid charm or maybe what u meant is this ?

[twin grail]

i got this for 2 years

[dry kestrel]

@west stirrup Please stop with the text wall, the cap is 6 months of Pro. https://www.perplexity.ai/help-center/en/articles/10964633-how-can-i-refer-friends-and-classmates Please don't share your private email on discord for safety reasons.

If you require any support you can email [e-mail]

[twin grail]

i thought it was 24 months

did they change it ?

Oh ye they changed it

rip

[valid charm]

I was asking for you to share the method for getting Pro for free in case you had one lol

[west stirrup]

thank you

[valid charm]

There is a guy on this server that has all the AI premium subscriptions for free. He never shares the methods though

[twin grail]

the batman guy ?

[valid charm]

When they launch a new top tier subscription feature we ask fhim to try it for us

[valid charm]

I think he goes by Joe

[twin grail]

oh ye

[sinful merlin]

3 business days. so am i going to have to do a string of emails?

[twin grail]

he shared me the pro method

for chatgpt

[valid charm]

WHATTT!!!???

[dry kestrel]

got abused

[valid charm]

lololololol

[sinful merlin]

i lose access to the email that i have the account on in 7 days. like if i don't get a response back, am i just finished?

[twin grail]

ye they were selling it all over the place

[twin grail]

U can send another email to them saying u will lost access ur email in 7 days and tell them to send u a email in another email of yours

should be all sorted out

[twin grail]

I love pplx max more than chatgpt pro