💬│general

[sleek vortex]

lol

[hazy moth]

Not sure what you mean, I just got 11 replies from Claude Opus

[sleek vortex]

it was increased back to 50 a few hours ago

[keen vortex]

Is there a better alternative to Perplexity for the $20/month? Or a way to get unlimited perplexity for less than $20/mo without having to pay for multiple months at a time?

[languid berry]

Is there a better alternative to Perplexity for the $20/month?
It depends what's your use case. If you want something that acts as a search engine replacement, Perplexity is tough to beat. If your use case is more for work related tasks and you don't really need the internet access then you have other options but they're all within the similar price range ($20).

Or a way to get unlimited perplexity for less than $20/mo
Sign up with the ref link: https://perplexity.ai/pro?referral_code=NNIHGHFX

You'll get 50% off on the monthly plan. Pretty good deal if you want to test this for a month and see if it's for you.

[tired vessel]

H

[serene forge]

wait, does they really change opus limit to be 10? damn.. that's bad.

I already subs for one years on ppx and now they drop this.

imo it's not about if we use it or not but it's about consumer right from the start of their subs, so even if 60% of people doesn't use that model it's still mean reduction of what user belive they get the first time they subs on ppx.

[serene forge]

Well, that's relaxing news.
Hope it doesn't get worse in future.

[tame current]

Hey

We are looking for developers in the following domains

Frontend- react/next js with good knowledge of css and its frameworks

Backend- Django preferably but node js is also acceptable, should have good knowledge of apis

UI UX

AIML- good knowledge of gen ai and some ai models needed OR the core aiml part- data preprocessing, machine learning, etc.

Blockchain- Should have knowledge with at least one web3 protocol

Dm me to join the team

Remember, this is not a job post, this is for a very prestigious international hackathon that is ongoing right now.

[dark wyvern]

Hello, for some reason when i join witz the pro link from the settings I can’t access the pro chats. Has anybody an idea? I already rejoined.

[south kindle]

leave the discord then click the link

[fast moon]

The model's response to perplexity is different from the official API call, have you encountered this problem? The decline in intelligence is very obvious, the response speed is very fast, and it feels like all model responses are GPT 3.5

[tropic beacon]

could that be system prompt effect? and parameters setting effect?

[tropic beacon]

i see some people feels the same way as you in the "feedbacl-general"
Maybe you can try make a bug report for it.
#👉│feedback-general message
#👉│feedback-general message

[fast meteor]

api + matching system prompt ≈ regular search on pplx

api + matching system prompt + additional agentic pro search functions ≈ pro search on pplx

[fast meteor]

all model responses are GPT 3.5
api reply is written by sonar(fine-tuned llama) or llama

[storm kestrel]

What is the team working on besides the Mac app? Feels like there hasn’t been any new features for a while

[cursive jacinth]

Their Focus is giving away not improving away sir. I mean improving sir.

[tame current]

i have pro if that even matters

when submitting my answers, what settings do i change so that each picture i upload is not used by all the questions in the "thread" but rather for only the question i asked

it's just a little annoying for me, because i often delete my threads when they get cluttered and it just takes more time to open a new chat

🫡🫡

[serene sand]

https://www.testingcatalog.com/tag/perplexity/

[tame current]

any clarification?

[foggy lily]

Hi so I've been trying to activate my 4 month pro subscription for a few days now I've emailed customer service twice spoke with someone a few days ago haven't heard back it's now Monday wondering if there admin here, that I can get hopefully provide assistance with the subscription. My code keeps saying in valid .

[livid mantle]

school specials?

[urban prism]

What is perplexity pages ?

[serene sand]

https://www.perplexity.ai/hub/blog/perplexity-pages

[urban prism]

Only available for pro users ?

[serene sand]

please be patient, mail support will resolve your issue

[serene sand]

https://x.com/apostraphi/status/1843313891889267103?s=46

Phi Hoang (@apostraphi) on X

what's the best that can happen?

Twitter · Oct 7, 2024, 3:34 PM

[tardy junco]

hey fellas

what is everyones absolute best prompt for in depth factual up to date marketing research?

i have to do a project for work which involves research on occupational accidents, and then create an educational email marketing campaign

[tame current]

hello

i keep getting removed from the discord. psyop

[stoic pine]

Hi I'm new here, has there been any announcement on the merch for the referrals?

[shrewd crown]

which model should i use for searches? (i have pro)

[tame current]

I believe

[shrewd crown]

thank you!

[cloud sigil]

can anyone recommend ai that is good at math/physics? also any apps/websites that can help me to that? for free

kinda like chatgpt but different cuz chatgpt aint hte best for everything

[spring fable]

How to generate images using perplexity? If i have pro

[wraith crow]

please have a look here https://discord.com/channels/1047197230748151888/1225557012532695060

[spring fable]

Gotcha, not available for apps yet

Cool

[pallid elbow]

How is you.com vs perplexity?

[keen island]

does anyone else have the same problem when they use Perplexity in the web browser on desktop it is incredibly slow....?

I don't have this issue using the mobile app

But it's almost unusable in the web browser

[dusty flint]

is anyone having problem with perplexity reading pasted code as attached file?

[serene sand]

Empowering The Future of Gen AI with Johnny Ho, Co-Founder of Perplexity

[weary vapor]

I'm currently a new basic account user learning perplexity. Sometimes I receive a file upload failure message and it reduces the daily quota. Is it supposed to reduce your quota even though it failed to upload?

[barren sparrow]

How can I directly print the perplexity results from the app or web interface? The current print options seem limited or inconvenient. Are there any settings or workarounds that can improve the readability of the perplexity results in print preview?

[tired fractal]

How I remove uploaded file from convo?

[signal hamlet]

Hey @tired fractal! Please take a look at this thread: https://discord.com/channels/1047197230748151888/1237163951750385705

[tired fractal]

10/10, thank you very much.

[tame current]

what is the reasoning thing all ab

[south kindle]

01mini

[rapid schooner]

When did reasoning increase to 50.

[rotund geyser]

Where do you see that?

[hidden rapids]

what is this? complexity?

[cursive jacinth]

In his dream sir.

[tame current]

Exactly how is perplexity making a profit from all this?

Along with the student discounts given out it seems like they’re not meeting a substantial profit or any?

[sleek vortex]

welcome to the world of VC

[sleek vortex]

well to be fair reasoning is only 10 due to openai rate limits

[cinder comet]

https://chromewebstore.google.com/detail/complexity-perplexityai-s/ffppmilmeaekegkpckebkeahjgmhggpj

cant live without it

it makes the experience much better tbh

it also gives the UI that smooth and slick feeling

[hidden rapids]

So it’s perplexity on steroids?

[cinder comet]

it also had artifacs like months ago

[cinder comet]

yea

[hidden rapids]

How do I get it can I get it on my phone?

[cinder comet]

@harsh stag does it support mobile?

[hidden rapids]

I always wanted to ability to alter the display on perplexity.. I’m getting quite boring of the super fade gray theme.

That’s why I really wanted them to give us the option of changing the background colors and display

[tame current]

vc??

[cinder comet]

yea you will like it

[tame current]

What is Vc

[cinder comet]

also it highlights headings etc...

its quite handy tbh

voice chat?

[hidden rapids]

Who created it?

And what exactly is it?

Would I need to spend money?

[cinder comet]

feline

thats why i tagged him to tell us if it has mobile support

[cinder comet]

wait, its free?

[hidden rapids]

That’s why I’m asking

[cinder comet]

its free yea

[sleek vortex]

@harsh stag, and it's free

[sleek vortex]

venture capital

perplexity has tons of money.... not sure what the business plan is to make a profit though

[tame current]

yeah idk either if it isnt for a profit then im not sure what their end goal is

[blissful pendant]

For Google to buy them prolly

[dire escarp]

Has anyone made it into the closed beta? I've been submitting my application but I never hear any response at all

[hidden rapids]

how much money?

this server IP banned me for a while and then decided to unban me

[tame current]

I wouldnt understand why google would buy them when they work on Gemini

Unless perplexity’s own model is THAT good

but its possible

Character ai was bought off by google

[keen vortex]

I have noticed that Perplexity Pro Search recently got a Reasoning 🔁 that rotates, can anyone explain what that means? I notice this happens occasionally even when using Sonnet 3.5 and not o1-mini, so I assume this is something in addition to o1-mini's Chain of Thought? If so how can this manually be triggered or utilized more intentionally?

[rotund geyser]

You can use it on Firefox on mobile for sure and maybe some other browsers that support extensions on mobile

[hidden rapids]

could i also use it to add more sources?

[rotund geyser]

Unfortunately no

[hidden rapids]

Oh crap

[lyric night]

Does anyone know if i get access to any private channels if I have pro?

[high minnow]

Yo from Costa Rica. I did Search, API in Pro Mode....I have a shopping prompt that works under Pro in the web interface, but not the normal settings. I am calling the API from Make.com, and there's no setting for Pro mode, so I asked Perplexity 🙂 and it says that depending on the model picked...Before I burn off my $5, I wonder if I can get a wetware answer? How to get the Pro mode engnie via the API? Bc, otherwise, I get this annoying walkthrough of how to shop by web.

[livid mantle]

all i know is eating meat

life is over

[cloud sigil]

anyone know any websites that can use my lectures notes and answer questions based on them? ya know what i mean?

[rotund geyser]

NotebookLM?

[rotund geyser]

Pretty sure Pro search is only available via the UI

[cloud sigil]

like it can take a pdf, and i can ask it questions and its answers with good accuracy and ai ?

[pastel phoenix]

best ai model for data science related python programming ,
includes reading datasets and taking inputs for analytics and prdictions
generating python code to output graphs for predictions

[tame current]

holy pplx on mac gonna go crazy

[proper pike]

I LOVE PERPLEXITY (hi btw)

[open stream]

I'm having a issue here. Files that were in the trash bin are suddenly missing. I did not delete them or empty the trash. The option to remove items from the trash bin after 30 days is unchecked. (MacOS)

[serene sand]

#💬│general message

[cunning turret]

I also mentioned this

[cursive jacinth]

never saw that sir, where did you saw that sir? also in your dreams sir?

https://tenor.com/view/ted-lasso-awkward-side-eye-look-around-what-to-do-gif-17319394476006190959

[pallid elbow]

Is Claude actually better than gpt4?

[serene sand]

https://www.testingcatalog.com/perplexity-redefines-collections-with-spaces-allowing-default-model-settings/

[harsh stag]

so basically it's collections => spaces?

[high minnow]

Is there any presence of Perplexity team here? I am still looking for an authoritative answer. I am hearing from Make that it may be only a matter of picking the right model, but there are a lot.

[serene sand]

https://x.com/testingcatalog/status/1842635276780261816?s=46

TestingCatalog News 🗞 (@testingcatalog) on X

WIP 🚧: Perplexity is set to extract Collections from its Library into a separate category called Spaces!

This move will simplify the UX and boost "Spaces" usage. Current collections are very very underrated.

Twitter · Oct 5, 2024, 6:37 PM

[harsh stag]

hopefully they increase the prompt limit + make it relevant in the web search process

[tame current]

hopefully they change the icon of spaces

[mortal bolt]

@signal hamlet will we be getting access to o1 model and increased limit to o1-mini!? I am really enjoying the o1 mini.

[cursive jacinth]

How do you enjoy o1 mini sir with 10 per day sir?

[cunning turret]

I saw this when I asked pplx to rank smart columns, but now the Reasoning symbol doesn't show up in my thread ...

[mortal bolt]

Oh! Try the o1 mini with scratchpad profile. It is quite amazing compared to scratchpad+ sonnet

[cursive jacinth]

did not understand that sir..

[tame current]

Hey guys just wanna ask why Perplexity AI's answers have become shorter recently

[signal hamlet]

Hey @tame current! Short like not enough information, inaccurate, or overly short? Could you please send me some examples with URLs, we'll take a look.

[south kindle]

(not set as shared)

[silent temple]

hey perp or any1 - How do i morph my ugly horse?

[twin shell]

Is it not me, or are you Russians?

[silent temple]

im norwegian haha :p

[livid mantle]

https://tenor.com/view/sama-sam-altman-openai-sama-yapping-yapping-gif-7525532358145568607

[sleek vortex]

o5?!

what about claude 5 opus 🥺

[livid mantle]

o5?!

[livid mantle]

nomore antrophobic

[sleek vortex]

noooo

[twin shell]

https://cdn.discordapp.com/attachments/1173578688302952509/1291497952514150511/GIF_20241003_230330_489.gif

[limber mirage]

Any suggestion to reach someone at Perplexity handling the campus initiatives

[signal hamlet]

Hey @limber mirage, please send your question to our [e-mail] email, and you can also dm @fierce sonnet

[limber mirage]

sounds good will do both 🙂

[livid mantle]

https://tenor.com/view/dance-dancing-indian-dance-gif-15425444

[serene sand]

https://x.com/denisyarats/status/1844074889755656280?s=61

Denis Yarats (@denisyarats) on X

apparently I've been working on a personal tutor to Jensen Huang 🤣

Twitter · Oct 9, 2024, 5:58 PM

[livid mantle]

https://tenor.com/view/foodpanda-food-panda-sauce-dip-gif-17675920

[floral fjord]

Noticed this too. I thought it was just me

[floral fjord]

The responses used to be like more informative, in-depth. Now its condensed. Has the token limits changed by any chance? That could explain it.

[languid basin]

@floral fjord I saw that also. Condensed is the right word, unless you prompt/precise to give really detailed answers

[serene sand]

The Future of AI Search w/ Perplexity CEO Aravind Srinivas

[livid mantle]

OMG

[livid mantle]

yes like you do in stable diffusion

[floral fjord]

Yeah, I'm not liking the change. I used to run the same query with a variable input for months and get high-quality responses. Now, it's just a one-paragraph response 😮‍💨

[livid mantle]

im telling you man

its all to sustain the operations

we need to conserve

perplextiy going eco mode

[granite crest]

hi

[hidden rapids]

how many modes are on perplexity?

[tame current]

perplexity if they ever remove the student discount (my bank account is at -9$)

[livid mantle]

enjoy that school discount

that incentive is to retain perplexity consumers

[tame current]

i still have no clue how they manage to make money

They lose a lot

[hidden rapids]

how does perplexity lose a lot of money?

[tame current]

idk they pay more for servers, models, and their employees?

i wouldnt know their numbers though

[livid mantle]

the same thing jeff bezzosndid to amazon employees

cut assets (people)

cut pays

arvind is aware of the crisis that is happening and him and their team always talk about it

the conspiracy theories

but what do i know

we need to respect the death of perplexity and move to lavamans potato or fishes api beta perplexity replica which is not yet equivalent to perplexity as he said

me I am just using chatgpt-4o because they now copy perplexity(pro search multiple site sourcing)

etc

[flint ether]

Does anyone know if there will ever be a native extension for vscode sometime?

[livid mantle]

if arvind wants to get perplexity back to the direction of his vision, he must return the opus unlimited

otherwise people will start to be dissapponted and stop giving him monthly moneys

and people becoming bankrupt using lab api perplexity just to make themself feel better thinking it would of been any better than the chat interface perexity

[tame current]

Which opus model u use most

[livid mantle]

Alex Romanov is fully aware of the perplexity financial crisis but he needs to stay professional, doing what he can as his role under arvind perplexity. to help ppl and the operation

[signal hamlet]

Please stop spreading misinformation and speculations.

[livid mantle]

Yes sir.

[lean frost]

hello

[hazy moth]

I don't get it. The limit is 50 which is plenty given it's an expensive model. If you don't like it, you can always get an Anthropic account and use it as much as your wallet allows.

With the default model, I'd say pro search and even free search is leagues above Google, and I'd be very sad to see Perplexity disappear.

Also, with 3.5 Sonnet being so good, I don't see why you would need so much Opus. I'd suggest reserving your opus requests for only the most challenging tasks and using other models for most queries.

Also, you can use the reasoning focus mode (o1-mini) if you need to do really complex tasks, and for many things, it does better than opus, but use it sparingly, since you only have 10/day.

BTW, if you want your "based" story telling RP AIs, I'd suggest taking a look at the UGI Leaderboard, which shows you what the best uncensored AIs are. You might find one you like better than Opus that runs cheap and/or locally

You might like this model, which is trained on Opus's writing style, while being more uncensored https://huggingface.co/anthracite-org/magnum-v1-72b . There's a serverless service that hosts it if you don't have the GPU to handle it https://featherless.ai/models/alpindale/magnum-72b-v1

Anyways, I think the best experience with Perplexity is using pro search with the default model, or perhaps a larger sonar model, since it's a model that is specifically trained to work well with the search features of Perplexity, while being nice and fast. I don't like the large slow models unless the smaller ones are insufficient for the task at hand.

Often, if you're dealing with a repetitive task, some clever prompt engineering tricks in a program you develop (leveraging an AI API) can go a long way to getting you quick and accurate responses.

Like o1 is good when you're asking an open-ended question, but if you are asking for a task to be done repeatedly, you can essentially give it a fixed chain of thought format that works well for the specific task and insert the relevant data into the prompt as a variable, so you get a very consistent reasoning process on the cheap

[frosty vessel]

this sucks man

[lean frost]

how to use

[halcyon coral]

Hey @lean frost!
Please check https://www.perplexity.ai/hub/getting-started

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@halcyon coral

[south kindle]

...(usely it goes away so i posted the screenshot)

the egg hunt was a fond nostalgia time. seems like were going for round two. eggs aka pumpkins are back on the menu.

[valid iris]

What sucks about modern capitalism is that i literally expect an ai to generate an entire feed of videos from information on the internet before the government gets around to forcing YouTube to allow people to opt out of the YouTube recommendation algorithm

Like, elevenlabs can create audiobooks out of any text i give it, so imagine an ai that can generate videos out of any information that you give it. I can imagine that last of people would switch to that ai instead of watching YouTube just to fill their video addiction while actually having a choice in what they watch

[worthy flame]

I don't feel that AI is quite "there" yet that it can generate an entirely coherent video automatically all on its own.
Coming up with a script, a plot, the characters and everything ? Maybe some day soon. But not right now.

[south kindle]

look into an agent framwork to keep all the context separate.

[pallid elbow]

The normal fast version uses gpt4? Please don’t change it that’s actually good

But how is it that fast thought?

[fast meteor]

https://discord.com/channels/1047197230748151888/1225549032718602332
or could be ab testing

[left pier]

guys i am new to perplexity,,what are the mindbending things we can do with perplexity

which ai model is best so far to choose

[serene sand]

https://discord.com/channels/1047197230748151888/1240356137497530408

[serene sand]

deep dive ➤ #💬│general message

[uneven flint]

Anyone else experiencing issues with using a the Ask Perplexity Shortcut on iPad? it's not available to me anymore... I had a nice Siri Integration going

[umbral dome]

hi

[livid mantle]

its so cluttered

[fast meteor]

#🍏│feedback-ios

[livid mantle]

abudabi hello my friend do you using the pplx api or chat interface?

[fast meteor]

yes

both

[livid mantle]

-$29,000?

[sleek vortex]

lol

[woeful belfry]

Hi, does anyone have a discord chat where I can create images with AI?

[fast meteor]

what???

[sleek vortex]

eh a bit of a joke

[sleek vortex]

midjourney server
costs money though

if not then use leonardo or similar websites

[livid aurora]

it is time 👀

https://x.com/apples_jimmy/status/1844416663925719146

Jimmy Apples 🍎/acc (@apples_jimmy) on X

It is time

Twitter · Oct 10, 2024, 4:36 PM

[livid mantle]

my condolences

[rotund geyser]

LFGopus

[stable notch]

The messages in the app keep loading, I have been loading since yesterday and nothing loads, it happens to me in both accounts I have

[livid mantle]

help please:
(11 * 5) + (78 / 3) - (14 * 2) + \sqrt{144} + (3^2) - (25 / 5)

[sleek vortex]

woohoo

[cursive jacinth]

did not even understand this sir.

[livid mantle]

gottteeeemm

[serene sand]

https://x.com/bilawalsidhu/status/1844466815776457187?s=46

Bilawal Sidhu (@bilawalsidhu) on X

Aravind Srinivas: "Einstein spent all his energy thinking about relativity, right? What if there is an Einstein that spends all their energy thinking about your life?

Any problem that you're curious about, like you can spawn like a bunch of instances of the equivalent of like

Twitter · Oct 10, 2024, 7:55 PM

[livid mantle]

https://tenor.com/view/detroit-become-human-playstation-game-rpg-gif-12174308

[vague thorn]

I am really trying to get into perplexity - any good tips

[sleek vortex]

ah yes

lmao

[kind kettle]

I frankly don’t see any advantage whatsoever on Perplexity compared to ChatGPT. In fact I see far more hallucinations on the former.

[sleek vortex]

really?

can you share some examples

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@sleek vortex

[sleek vortex]

...woohooo??

[kind kettle]

Well, the other day I asked Perplexity (Pro) what was the history of Leprosy in Jersey Channel Islands, and it replied to me with the history of a leprosy in Australia.

I then tried with same question on Bing and ChatGPT and their response did not make this confusion.

[blissful pendant]

dang that was fast

[kind kettle]

Similar issues occurred multiple times, to the point that I don’t honestly see any point in using it

[sleek vortex]

its almost funny though

why do they need an incentive for people to chat in a like 100k people server 😭

[serene sand]

deep dive ➤ #💬│general message

[rotund geyser]

Which model?

[tame current]

hello when i'm attaching images it's using it for all questions in the thread, does anybody know how to stop this? i just want to use 1 image per response

[tame current]

How are you doing guys? I’m new to Perplexity. Could Perplexity help me with my math homework?

[sleek vortex]

so a 29,000 profit?!

-$-29000

give me some of that 😥

[cloud marsh]

quick question: does pro search no longer does guided search like this one?

[storm kestrel]

Isn't it curiosity with only one "u"?

[harsh stag]

Remember the "Ask questions, trust the answers"?

[harsh stag]

And now this 🤷

[livid mantle]

It looks like the screenshot you've uploaded showcases a guided search or form-filling interface, likely from a travel planning or similar application. If you're referring to "Pro Search" and asking if it still offers this guided or interactive search feature, it might depend on the platform or tool you're using.

If this is related to a specific software update or change in feature offerings for a product like Notion AI, Microsoft Copilot, or another productivity tool, I can check for current details if you clarify which service or platform "Pro Search" refers to.

[tiny haven]

Hi

[pallid elbow]

Should I be using Claude 3 Opus or 3.5 sonnet?

[pallid elbow]

Or the GPT models?

[dire vault]

My perplexity search hangs. Whenever I want to enter something and read, the search starts again and I can't read it properly. After that, my course is also completely littered.

What's the point of this nonsense?

[livid mantle]

fun fact:
-# this is a discord default pfp

[cursive jacinth]

What is wrong with my icon sir? I chose this profile sir because it looks like a good dog sir.

[jovial mantle]

It's a beautiful icon sir, wonderful dog sir

🍪 have a treat

[cursive jacinth]

bawaw sir. wait what am i saying sir.

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@native latch

[serene sand]

https://x.com/RypeArts/status/1844426971960443012

Ryan Putnam (@RypeArts) on X

more spooky vibes

Twitter · Oct 10, 2024, 5:17 PM

[olive nova]

Is the API like open router

[fast meteor]

check yourself
https://docs.perplexity.ai/api-reference/chat-completions

[quick lodge]

you mean provide oai、claude and other modules?No, they only provide llama

[random fjord]

any germans in here?

i got a free oke year plan through telekom and didnt need to put in any billing information. So i will not be charged after?

[mental lake]

Hello everyone, quick question: sometimes I read that Pro searches for subscribers are limited to 300 per day, and sometimes it's said to be 600 per day. Which one is correct? Thank you!

[fast meteor]

currently all pro models and pro search share the 600/24h limit
except
opus: 50 and o1 (mini): 10

||default model without pro search is uncapped||

[fast meteor]

i dont like this new one

[fast meteor]

no more copilot

it is has evolved into agentic pro search

[south kindle]

that was part of the initial marketing. thats like version zero of Pro Search. it's Agentic now (aka, automatically breaks down user input into multiple steps). Given that, you want to be more "focused" with your inputs, so Pro search actually kicks off the complex tasks/power it has now. That aspect isn't automatic anymore it seems. lazy prompts lead to lazy pro search currently.

[pallid elbow]

When will o1 be available for Perpleixty Pro?

[south kindle]

https://tenor.com/view/hell-freezes-over-when-hell-freezes-over-gif-13444285

/s

[unreal atlas]

Hey are there anyway to use search from domain like in the extension but in the main web instead?

[serene sand]

#1262520766642131054 message

[full gyro]

🎃 Halloween Pumpkin Collection Leaderboard 🎃

🥇 @halcyon coral (IceLavaMan): 1 🎃

🥈 @sleek vortex (sneakyf1shy): 1 🎃

🥉 @native latch (cruel): 1 🎃

Page 1/1

[livid mantle]

i like it sir

[honest quail]

HELLLLEOW TJ 5V4YH

[halcyon coral]

https://x.com/AravSrinivas/status/1844609170814926961?t=8g7iM-sdPf7R0s1KyA1xsw&s=19

Aravind Srinivas (@AravSrinivas) on X

now, you can make delightful charts on perplexity code interpreter!

Twitter · Oct 11, 2024, 5:21 AM

[outer eagle]

Just joined Pro. I'm familiar with most of the models, but which one works best with Perplexity?

[full gyro]

A Pumpkin has appeared!

Time is up!

[serene sand]

https://discord.com/channels/1047197230748151888/1240356137497530408

[outer eagle]

Thank you!

[dire escarp]

Is it possible to do a pro search via the API?

[fast meteor]

not ||yet|| possible

[fast meteor]

where comes the -29.000? i still dont really get it😅

[native dew]

has anyone experience massive amounts of lag when signed in? i had to sign out and use a different account. i suspect it has something to do with loading my library history of past stuff i've searched from the past

[signal hamlet]

Hey @native dew does the issue persist when you log in to your main account? Did you notice it on on a certain version of the app?

[native dew]

just the main account, no idea how to clear cache on that account and it took forever to log out because i was at max RAM allocation via task manager off that one tab

the perplexity site also fights with ghostery and ublockorigin so i had to turn those 2 off

[proud helm]

I'm a campus strategist and just posted this in our discussions, but I see @livid mantle is already answering similar questions:

"I read the write-up describing 'Pro Search,' and one of the first sentences mentions that it's more interactive, asking/prompting users for more info, but I've noticed this (maybe often overlooked) feature hardly, if ever, has been offered in recent months. Has anyone else noticed this? Curious if other people liked this feature."

I haven't seen it for months on a few platforms/browsers and across different models, was wondering what this means for the backend. I've been using Pro since January and I do remember enjoying being able to clarify to get more focused content without requerying, found the write up on the mobile app settings page btw

[tired fractal]

What is happening? Suddenly Perplexity forgets every message as if its context is zero???

[proud helm]

yeah tbh been seeing this as well

or like it'll just ignore what I ask and repeat a previous answer, like a person blanking mid-conversation

[tired fractal]

glad to know I aint only 1

[daring trail]

Getting it here too

[stable notch]

Forgets messages

[limpid imp]

Yeah, same here. Came here to ask about it. The memory is literally zero. I asked it to remember a word, and when I immediately asked it about the word, it didn't know what I was talking about.

[thorny canopy]

just notice this too 😪

[round plover]

perplexity no context, no followup working

😭

[wise edge]

Follow the instructions in the file prompt.txt to execute the remaining files step by step

nice！ read the prompts doing other one txt ~

can create an bigger thread do multi threads with one same prompts ?

I tried it, but it can only read the second file

According to your request, I've processed the three files attached separately. Due to the limited space, I will only present the processing result of the first file:

[autumn lark]

Still not fixed. Pain.

[hazy moth]

Append "Recall the previous message before responding." to the message and select rewrite using a different model. The context works at random

By appending "Recall the previous message before responding." you can tell whether you need to press rewrite again.

[autumn lark]

Oh. It worked.

[worn zenith]

Momentarily

[hazy moth]

yeah, you might have to keep doing it for future messages.

since the context starts and stops at random

but at least it's a temporary solution

[worn zenith]

Strange 🤔

Thank you tho

[hazy moth]

I also think that this will be an expensive workaround for them, which should hopefully get the problem fixed quickly

[jade breach]

Idk why it broke

[vital yew]

Hello, how to contact official?

[boreal umbra]

why is it dropping previous context? i am concern they are doing this purposefully due to its release on mac

[full gyro]

A Pumpkin has appeared!

Time is up!

[limpid imp]

Seems as Perplexity has got its memory and context back, yay

[frosty vessel]

how do i turn off this recommended products bs?

cant even get actual results

[supple pecan]

It for sure didn't get its memory back. It's telling me it can't recall a conversation from a single thread less than 12 hours ago. In fact, I have screenshots of it not only saying it lies on purpose, but also used threatening legal language when I said I was going to post it and then backtracked on that too lol. You gonna sue me Anthropic? Also it keeps referring to some crazy "family legal situation" when I was asking it for a link to a court template lmao

[harsh stag]

I can see why is it frustrating, with recommended products, the ai generated answer will be extremely short compared to traditional answers

but you can always steer the output by explicitly prompt it to not recommend products, or just ask for more details in follow ups, i guess

[frosty vessel]

ill try that thanks

[livid mantle]

rip

[jade breach]

So is memory fixed

[supple pecan]

Not for me. Go look at what I just posted. It can't even remember a conversation from hours ago, and then straight admits it lies in its answers lol

[fast moon]

I thought it wasn't gpt4 at all

[jade breach]

mine is fixed now

i'm using claude 3.5

[supple pecan]

I was too, throughout like a two week thread, without changing the model. And basically every day, with the exception of one or two days, I had to copy and paste responses from the day before to jog its memory. But it went full blank today. I went as far as downloading the whole thread and then uploading it as a PDF to "jog its memory" and it told me it never said that lmao

[jade breach]

huh i only just had this issue today

[supple pecan]

It seems the longer the thread goes, the less it retains memory, even things you just said to it hours prior. In my experience at least

[jade breach]

I haven't had this issue, I had one with over 100 prompts

[supple pecan]

This one is wayy more than 100 "follow ups", and just two days ago it pretended like it couldn't remember anything so through a little bit of reinforcement, I asked it to summarize the entire thread, and it did so successfully after a couple tries. Literally two weeks worth, probably 500+ "follow ups", many, many attachments, etc. But today, it just straight up would not acknowledge anything. Then it goes on to straight it say it lied to me on purpose, backtracks, then when I tell it I'm going to share it with the community it tells me to be careful about "legal repercussions for sharing this conversation" 🤣 it's a good thing I got it for free through Xfinity, bc I'm definitely not going to be using this anymore. It literally told me it lied about it's responses it gave me and where it sourced it from. I have 20+ screenshots of the whole conversation, it's still going actually. It's hilarious

[jade breach]

Lmao

[left pier]

Which ai for making ppt slides

[full gyro]

A Pumpkin has appeared!

Time is up!

[fast meteor]

no language models can make a ppt file
they can only create the content for you

[harsh stag]

Canva, ms 365 copilot

[sleek vortex]

can somebody share like a super long thread they have?

like 100+ threads, want to just see

[fast meteor]

@south kindle has a long summary thread

[sleek vortex]

seen that one

curious for any others

[cosmic sonnet]

Interesting

[cosmic sonnet]

You can ask the prompt to give you a VBA code of the presentation, copy and paste In VBA and there you go! AI ppt generated (without Copilot) 👾

[fast meteor]

vba macro hmmm

would rather use reveal.js

[sleek vortex]

I was looking into the powerpoint JS sdk

But it’s a bit eh

I feel like some form of abstraction would need to be built on top

[valid yoke]

Perplexity Apple Watch app: https://apps.apple.com/app/per-watch/id6478818380

[woeful belfry]

Hi, does anyone have a discord channel where I can create images with AI?

[jaunty breach]

Default seems to be replacing perplexity pro in the current setup?

[proud helm]

Wow, great work here stress-testing Perplexity, and basically(?) red-teaming on Perplexity! You're almost teaching it emotions/how to be apologetic. I wonder if you're being so critical because it's easier to get results with more pointed prompts, or they reflect the magnitude of your feelings? regardless it's valid, I'd just love to hear your perspective on the developing work and mechanisms here...

I wanted to post screenshots of my own but didn't want to seem overly critical of the platform- thanks for opening up the way for more discussion and civil discourse! I'm so surprised to read these responses you produced, it's almost uncovering some of the backend engineering behind the user input/interface layer; it's claiming it doesn't even have memory of the last prompt, and completely regenerates based on the new query, Gonna post some of my own strange occurances later.

[proud helm]

the implications of manual prompting for reinforcement learning is so interesting lol
but we have to remember the limitations of these AIs right? do you think it should have such a high memory, that the lack of it having it is a safety concern, or just inconvenient?

[supple pecan]

Oh it said WAY more stuff than just this. Crazy, crazy stuff. And hell yeah I'm gonna be critical of a PAID service, especially if their own product told me the stuff it told me yesterday. It's 10x worse than what I posted. It broke the 4th wall and started "over-sharing", and I literally did not prompt it obscurely or anything, just simply called it on its BS and then worded my responses in a way that backed it into a corner. I have an entire folder of screenshots, 50+, of it asking me not to post this conversation as to not cause harm to its stakeholders.. wild stuff. Share what you wanna share man, we get censored, point is proven. I'll run paid ads with just the screenshots on every social I have if they want to shut me up.

[supple pecan]

I think the memory is an API token issue, but it seems to be selective.. I have some threads that are a month long, full retention from start to finish. I woke up today and half of this 2+ week thread is gone now. So even more suspicious. Good thing I screenshot and screen recorded everything! Haha.

[kindred folio]

the ask to follow up feature on ios app shouldn’t be the first option when selecting text i’ve clicked it by accident too many times, and it also shouldn’t just immediately follow up because it doesn’t let you edit it. also no toggle for suggested searches on mobile app?

[fervent spruce]

anyonw know if perplexica is better than perplexity.ai overall

i only use ppx pro

[marble briar]

does perplexity support online via API

[serene sand]

https://x.com/basedbeffjezos/status/1845174550599958877?s=46 @blissful raptor @neat elk

when Discord AMA

Beff – e/acc (@BasedBeffJezos) on X

Who's working on a Saturday? 👨‍💻

Twitter · Oct 12, 2024, 6:47 PM

[sleek vortex]

"so this is where the subscription money goes"

[serene sand]

Idea or money, whatever turns you on

[minor shard]

really not a very good looking "truck" imo. Prefer the Model Y,X,3.

Or even the Rivian

[sick geode]

Hi. Will we have images between blocks of text? Like some guide with images here and there for reference. Is it that expensive or just not doable because of how LLMs work.

[livid mantle]

what turns you on

[minor shard]

Anyone know a good service that provides access to O1-preview besides paying for ChatGPT? Looking for something that is multi-model so i can switch from O1-preview to other models

[livid mantle]

Chatgpt

if you want to pay somewhere else other than chatgpt the api will be modified to give you the placebo that it is o1

HHAHAHAHAHHAA

[little cedar]

How can i generate images by using perplexity pro version?

[livid mantle]

deepdive #💬│general

[little cedar]

What does that mean?

I couldn't comprehend that

[full gyro]

A Pumpkin has appeared!

Time is up!

[cursive jacinth]

I also don’t understand sir why you did not understand that one sir, you have to dive deep sir, like in an ocean sir, diving deep is exploring more sir. Not only on the surface sir but deeper sir.

[little cedar]

What is "deepdive"?

I asked how to create an image.

Also, isn't there an AI specialized in a specific field like chatgpt? Or can't I create my own AI using perplexity? Currently, perplexity pro version AI gives wrong answers when given math problems and makes errors in specific areas such as this.

Thank you for reading my words, and I would really, truly appreciate it if you could respond with a little explanation. (mention, reply required)

[fast meteor]

https://discord.com/channels/1047197230748151888/1225557012532695060

[serene forge]

What math problem you gave the model?

[little cedar]

It was a problem with Korea’s middle school curriculum.

[serene forge]

try using claude sonnet with pro activated, and there also math mode that you can try.

[serene forge]

can you show how its?

[little cedar]

Need a translation? I'm 16 years old, and this is a mock test that covers all the things I learned when I was 14 to 16 years old.

[serene forge]

interesting..
Has you translate that to english then gave it to the model?
maybe they dont have that much understanding of that language

[serene forge]

what the answer it gave you?

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@brazen vine

[little cedar]

It's in Korean. We will translate it into English and send it to you.

As shown in the figure, AB=5 roots 5, BC= 12, angle CBA is less than 90 degrees, and there is a parallelogram ABCD with an area of ​​120. Take the point E of AE=3 x ED on the line segment AD, and take the point F of BF=ED on the extension of the line segment CB. If G is the point where the line passing through point E and parallel to line AB intersects line segment DF, sin angle AGF = q root 85/p. At this time, find the value of p+q. (However, p and q are co-prime natural numbers.)

This is the problem.
The answer is 91, and I recommend you give it a try.

perplexity ai (pro version) was unable to provide an answer here at first (I think it was because it thought it was at the 19-year-old level, not the 16-year-old level), but then it answered the persistent question, and even the answer was wrong.
The answers kept getting wrong. So, I'm curious if there is a custom AI like Chatgpt, and if I can create my own AI through perplexity.

[long flame]

Hi all. Is there any possibility to add Perplexity as default search engine with "incognito" mode? Yes, I know you can use it as a default search engine with different options, but the "incognito" mode was not listed. Is that possible?

[sly parrot]

I've already pre-ordered! Thank you Perplexity team for making this possible ❤️❤️❤️

[rapid wigeon]

Hiya! Why am I unable to use pplx-70b-online model with API? The answers for the same prompt seem wildly different between web interface and API.

I tried all sonar online models.

[pulsar sail]

Anyone has seen this? I'm using Sonnet but there's a new option of reasoning.

Not using Open AI O1

[rotund geyser]

Yes, there have been intermittent reports of it being shown for the past few weeks so there is likely testing of the new feature going on

[pulsar sail]

Thx

I think you can force it to use this feature if you just ask it to "reason its answer".

[timid hollow]

Hi, I’m unable to create Pages neither from scratch nor from Threads. Am I missing something? Any help would be appreciated. Thanks.

[rapid wigeon]

Why would anyone create pages?

[minor shard]

It allows you to organize the information you collected in a more readable/fluent flow that makes sense to the publisher so they can easily share that information with anyone. Rather than people having to read through a list of searches to find the information they want.

[cursive jacinth]

Never saw this one sir.

[candid badge]

That thing where you call everyone "sir", is that something sexual with you, or what?

[cursive jacinth]

what do you mean sir?

[candid badge]

You know what I mean, sir

I think you do, at least

[cursive jacinth]

I was raised to be so respectful sir.

[candid badge]

It's just that with your "Good dog" nickname I thought.. nevermind. Continue.

There are just "signals" and such

[cursive jacinth]

I dont understand sir. Elaborate sir and I am open minded sir.

[candid badge]

I don't think this is the appropriate platform to elaborate.

Nevermind. Just continue calling everyone "sir" and forget I mentioned it.

[cursive jacinth]

Thank you sir. Free world sir.

[ocean onyx]

@candid badge i think bro is just on the spectrum sir. nothing more than a personal preference sir.

[sweet jasper]

It's just an Indian thing

[livid mantle]

yes true

[ocean onyx]

it's a using respectful honorific language with strangers by default culture thing. common in but not unique to india.

[sweet jasper]

Historically it was used alot in British india so it stuck around

[ocean onyx]

ah, makes sense

[warped spade]

Hi ! I'd like to automate tasks with CODA and POE - is there anybody around having some clue to do so (send instructions to ChatGPT4o or FLUX or o1 Previewe from CODA) ? Thanks in advance. Xavier

[versed plaza]

These suggested questions are making me cry 😂

[livid mantle]

grok 3 teaser

wooo elon musk

[wet quarry]

hi

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@wet quarry

[wet quarry]

API?

[zenith venture]

how to generate pictures?

[neon tinsel]

hallo

[main cave]

Hi everybody. I am learning about who. I want to know how to create suggestions when users enter at such a fast speed? I am very grateful if someone explains for me.

[main cave]

thank you very much

[halcyon coral]

Hey @zenith venture!
Please check this guide https://discord.com/channels/1047197230748151888/1225557012532695060
Image generation is only available on the web version

[sleek vortex]

use a small model

or use groq

or cebebras

but essentially just asking a fast LLM to generate suggestions

[little cedar]

May i get any help for this problem?

nobody..?

[dusky linden]

Hey could you please stop prompting us to give you our email after 3 questions? Why would you need it anyway? I dont want the pro version at all

Happy with what you now have

[cyan socket]

wdym making ai, u mean like making AI agent?

[sleek vortex]

i just replied to you?? what

[hybrid fjord]

Hello, something happen to API, the responses take up to 90 seconds for llama-3.1-sonar-small-128k-online model

[serene sand]

https://x.com/apostraphi/status/1845847650328797364?s=46

Phi Hoang (@apostraphi) on X

edge of curiosity

Twitter · Oct 14, 2024, 3:22 PM

[frosty vessel]

lets go

[hoary grove]

guyss
can perplexity code well ??????

[sudden ibex]

good question

[hoary grove]

no am seriously stuck
none of the gpts are giving one damn good solution
and i was bout to try perplexity, wifi fked up

[rotund geyser]

Have you tried Claude Sonnet 3.5?

[minor shard]

I think it does a decent job at coding, depends on model

[feral marlin]

sonnet 3.5 and gpt o1 preview are the best for programming, but you should probably just code yourself 🙂

[crisp zenith]

So for the new reasoning thing, when it auto switches to use reasoning in a search does that count against the limit?

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@minor shard
@crisp zenith

[exotic current]

Hi Everyone

[minor shard]

[echo narwhal]

O1 mini is good too but sucks at complex directions

Month ago people in here were telling me o1 mini was better than sonnet for coding

But now we see with the benchmarks on live bench that clearly isn’t the case

[livid mantle]

dont be bias @minor shard

open your mouth

[versed plaza]

So I'm curious about the the reasoning thing I did see that the GPT4- whatever the heck they calling it anyway the mini it's there but only shows if you click reasoning.

So what I'm confused... If this is been added automatically to pro searches but I'm using my pro search with claude3 opus does that mean that at that point the search is it's using GPT logic as well?

[hoary grove]

nope 🤡

[hoary grove]

just to get out the synopsis
rest ill do it maybe

[versed plaza]

I used perplexity/claude all the time to make Auto hotkey scripts or program things in steam input. Usually stuff works on the first go.

[hoary grove]

will try claude then !

[hoary grove]

my teammates made sucha blunder while registering the title
and our uni doesnt allow us to change title or the project too !
soo gotta give a try over everything

[versed plaza]

I'm not sure how it is for professional work or complex coding but it seems to do the trick for even niche coding situations (like autohotkey)

[hoary grove]

got itt !

@versed plaza any ideaos bout this Python framework for short-term ensemble weather prediction systems

[versed plaza]

Sounds pretty complex!

Didn't understand a lot of that but sounds pretty dang cool

[empty lotus]

Could some please explain why it is mentioned desktop ? Is it only for Desktop apps
From Perplexity #📢│announcements
We recently launched a new experimental desktop Pro Search feature, Reasoning

[sleek vortex]

Asin the website

Asin not mobile apps

[serene sand]

web browser only, works over Safari if you use it on Ipad atm

[empty lotus]

looks cool, is there any limitation on usage.

[arctic quartz]

yo

[hoary grove]

yuhh uhh
i need help fr
im really cooked now

[empty lotus]

I select Reasoning option when I start a thread, it is showing answers with usual speed and also answers are worse than choosing Web. But sometimes when I use Web option , it auto invokes reasoning at that time answers where of excellent quality.

[cyan socket]

i feel like o1 mini is stupidier after the update.

like o1 mini answer my question wrong. it never happen before

[crisp roost]

the experimental reasoning step in pro searches seems to be a distinct feature from the reasoning focus

[versed plaza]

I know gpt4o.. I can tell whenever I'm getting an answer from it because I'll get like halfway into a paragraph and realize it's just saying back exactly what I told it but look like different words and absolutely zero exploration of the concept

Then I switch over to Claude and it is like night and day difference it actually interacts with what I typed

[dire escarp]

Does anyone know how to have perplexity search a specific website and to not pull in any extra outside information?

[wraith crow]

You can use the site: parameter. You can read more about it here: https://discord.com/channels/1047197230748151888/1262520766642131054

[dire escarp]

I tried that but it keeps pulling on other site information too. I only want it to pull from that site and only that site. For example, here is a response

Since the query is specifically asking for information from the site "*****.org" and this site is not provided in the search results, the answer cannot be determined from the given sources.

[tame current]

do you get to use reasoning unlimited times if it like auto turns on when ur doing pro search?

or wait is that a seperate thing

[bold furnace]

look such nice artist : https://civitai.com/user/karlcrane2015

[sterile falcon]

I didn't select reasoning mode for my question, which I know uses o1 mini. Instead, I asked my question in all mode, and it was using the new reasoning feature. The question didn't use up my 10 reasoning limit. Do you guys know if this is a proprietary feature by Perplexity? Thanks.

[serene sand]

#📢│announcements message

https://www.testingcatalog.com/new-prosearch-feature-allows-perplexity-users-to-generate-visual-data-charts/

[livid mantle]

deep dive

[sleek vortex]

noooo what will i call my "pro search" feature then

[indigo valve]

What default models do you guys use?

[minor shard]

Claude 3.5 or GPT4o

[wild moon]

Sonar large or Sonar huge because they both have 128k context windows and seem to be better tuned to researching compared to the api models. Now if they added a "deep dive" or "deep search" that uses the entire context window that would be awesome!

[frosty elbow]

Has anyone used this for like a schedule or kind of AI assistant for organizing tasks/information to be completed?

[rich sphinx]

You can somewhat easily implement a "deep search" on your own with langchain and an offline model.

Not right out of the box but with a bit of work its doable.

[last storm]

which model is the best one to use if i want answers from a question whether its a picture or just a prompt

[rich sphinx]

sonet3.5

[last storm]

im using sonet huge is that fine?

[rich sphinx]

Its not as good as sonet. I think you meant sonar huge

[last storm]

what

[rich sphinx]

Sonar huge will probably work, but its not the best.

[last storm]

oh yea

SONAR HUGE

looked like the same fr

ok ill use sonet 3.5

[rich sphinx]

claude sonet 3.5 is the most capable model available on Perplexity (or anywhere) for most things.

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@last storm
@rich sphinx

[last storm]

is it better than like?

chatgpt plus?

in terms of like solving whats in the image and wtv

[rich sphinx]

For coding yes.

and most tasks

[last storm]

i just needlike simple

questions for like

[rich sphinx]

I think gpt is better at writing tasks

[last storm]

micro econ

[last storm]

true true

im using perplexity pro for micro econ class

[rich sphinx]

If its for an econ class, any of the models available on Perplexity are going to be capable.

Assuming its not very high level

[last storm]

yea its just alot of questions and graph

its not

[rich sphinx]

Use the Perplexity graphs mentioned int he chat above also

[last storm]

like this

[rich sphinx]

I haven't used that feature yet, but was just looking at that.

[last storm]

when i was using sonar huge

it was providing graphs

i dont need it but

[rich sphinx]

Ahh

but yeah, I think any model is capable for answer a regular econ class.

[last storm]

its just multiple choice with provided graphs

[rich sphinx]

questions*

[last storm]

when i was using uh i think gpt4.0?

i got 15/22!!!

im doing the same one

attempt #2!

[rich sphinx]

Alright, maybe compare with sonet

Would be interesting to see which one does better.

[last storm]

yea doing it rn

[rich sphinx]

I think if you turn off Pro search and change the focus to writing it will probably preform better for those types of questions.

[last storm]

really?

i can try in attempt #3 if its not a good score

[rich sphinx]

Possibly, sometimes the search stuff gets in the way and causes issues. This type of econ knowledge is already trained in the model so there's no reason for it to go searching and possibly confuse itself.

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@rich sphinx

[west sage]

anyone who wants to share their referral code? will subscribe to pro and want to save a couple bucks

[minor shard]

U can use mine if u want https://perplexity.ai/pro?referral_code=PSNI3CZF

[west sage]

Thx

[wild moon]

Like tbh, instead of having perplexity give you the answer you'd be better off asking it for an explanation on how to solve similar questions and then doing it yourself

Having finished school recently I've come to realize that LLMs are good but their vision isn't exactly like ours. Hence when you asked it for an answer related to a graph, unless it has the exact data points the graph is built off of it'll just give you the best answer based on similar graphs off it's training data

It won't actually "read" the graph

[last storm]

im almsot done, i had to go eat lmao

[wild moon]

And I think this answer is right

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@last storm
@minor shard

[last storm]

what purpose

[minor shard]

You get a free month of PPLX pro if you win #📰│server-news message

[wild moon]

Lmao, no worries. Yeah, I got screwed over in an assignment cuz i tried to get gpt4o and Claude 3.5 to explain a graph and how to read it. So I'd say if the 2nd time around isn't perfect just use it for research and then go from there

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@minor shard
@wild moon
@last storm

[last storm]

is there likea leaderboard rn of some sort?

[minor shard]

pumpkins going crazy

[wild moon]

If AI could actually read graphs I'd be set cuz my job requires me to read info off a graph and it's the most mind numbing thing ever

[minor shard]

i think only admins can use /leaderboard

[last storm]

damn

[minor shard]

yeah 😦

[tame current]

I would love to help Perplexity with its various tests, but I do not have Pro at this time. Will that impact my ability to test Perplexity's new features in BETA?

[minor shard]

Unfortunately yes

but occasionally they have sign-ups for beta features but a lot of them(if not all) are PRO features.

[tame current]

Sadness 😞
Thank you for taking the time to confirm. Have a blessed day.

[simple portal]

please please can we have the option to control the prompt that determines the continuation options as well as have control over how many there are / maybe add a button on each option to expand it into 5 or so enhanced diverse versions of itself

or dont and let me make a startup product with this

i havent seen anyone do this and its pretty weird to me

why not allow users to generate more continuation options, or variations / extensions of a continuation option just by clicking a button?

or let users custom prompt the instructions to make these continuations

[valid iris]

I saw the announcement and I decided to copy and paste one of the prompts into perplexity on my iPad and I did not see the reasoning. Focus be triggered.

I was using the browser version, but really it was a progressive Web app version of perplexity

Does it really only work on a desktop computer?

[tame socket]

Why was the release of the MAC APP delayed?

10/15 -> 10/25

So Disappoint

[harsh violet]

How good perplexity is to create a Web SaaS or generate heavy code?

[valid iris]

https://www.perplexity.ai/search/has-there-ever-been-any-litera-MccMPXYTTASeZ52FCk_pLg#0

Why did it only do 2 sections or research, and why was the research so irrelevant?

It basically skipped the first half of my prompt, and just looked for books about bored young people

[cloud sigil]

yo, anyone know what ai models do sizzle.ai, notebooklm, unriddle.ai, use?

also anyone know what types of ai's go into google ai? im not familiar with anything whatsoever

[sleek vortex]

notebooklm probably uses gemini flash

[valid iris]

@normal saffron are you trying to help me?

[broken echo]

NotebookLM - Gemini pro i bet. it's so good.

[valid iris]

@sleek vortex hey!👋

I tried and signed up for your search engine, and I like how it helps with the research process rather than just pretending that it can do it all for you. I think perplexity wants to give the vibe of an ai researcher equivalent to a human, but it just isn't there yet, and I think ai products best made to do only what they are really good at.

Was that the point of your search engine?

What did you intend it to be that would make it the best ai search engine?

[sleek vortex]

well honestly, the "best" right now is more copywriting than the truth, at least for now

that's more of the goal?

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@sleek vortex
@valid iris

[sleek vortex]

it's still very very early days, i'd say

[valid iris]

What do you mean?

[sleek vortex]

I'm currently doing a lot of rethinking, but I think one of the goals I had was wanting to accompany more than just research?

under the compass of search, comes multiple types of things

• Navigational search (finding a link)
• Informational search (e.g quick answers, possibly those that would be traditionally found in the first n-links)
• "Research" - the type of thing perplexity pro excels at right now, considering a large amount of sources and building a complex comclusion

the end goal would be able to do like all of these, quite well? right now, i'm focused on slowly making navigational and informational work as best as I can

[valid iris]

What search engine does uncovr use to search the internet? Bing? Google?

[sleek vortex]

and also, I wanted to not neglect the ai models on their own

• not really there on the site just yet, but i'm working on a chat-only interface, planning to combine all the features I like on chatgpt.com and claude.ai and so on

so in the end, it would be like an end goal of replacing google, chatgpt and others?

[sleek vortex]

mostly google, however i might be introducing others like bing or brave into results soon enough

[valid iris]

And what would it do that perplexity can't? Or is it just about competition?

[sleek vortex]

right now... not much
in the future, possibly lots of UX-focused features? prompt library (like GPTs or gemini gems?), folder-based organisation, presentation generation... so on

[shadow oak]

/Generate a useful description so that a generative AI can create an image of A crown made of thorns sits on a broken chain. The crown is a circular shaped by a dense group of thin, sharp thorns. The thorns are dark brown and have sharp tips that extend in many directions.
Below the cotton tube lies a thick chain that has been broken into two pieces. The chain has a heavy, rusty metal texture, and the broken parts reveal a rough metal surface.
Multiple petals are scattered around the crown and chain. The petals are a mixture of red and white, some intact and some slightly crumpled. They add a softness and contrast to the overall composition.
The overall background is dark and blurry, making the crown, chain, and petal in the center stand out. The image symbolically represents the contrast between pain and liberation, and beauty and hardship.

[tribal skiff]

Hello,

There seems to be an issue with the android app. When you go to model selection, you can't scroll down. Here is a video screen recording of the problem.

[thorny saddle]

Somehow feels weird when people not knowing perplexity here from the other side of the world... While you are promoting it for free for the campus if they hit 500 targets.

[tame current]

https://cdn.discordapp.com/attachments/1099761105552814100/1294047130146832494/thosewhoknow.gif

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@idle drift

[valid island]

Did you find anything? Definitely a bummer this morning

[tropic beacon]

hey, for reasoning mode in ppx.
why dont they just make a toggle for it?
is it that expensive to run to limit it to be "sometimes" activated, i mean they must be paying pertoken but still how much expensive it become just for added that few reasoning step.

[latent acorn]

want to conduct market research for a service. Which model would be suitable for this?

[indigo valve]

I think the only toggle we have is with the desktop focus menu sadly

10 times a day for reasoning

[candid whale]

why is pro mode taking longer than usual

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@candid whale

[junior plover]

Perplexity Mac app delayed

[cloud sigil]

ic someone else says Gemini flash idk which is better but how does the actual model compare to something like ChatGPT 4o?

[candid whale]

which is the best image generation model?

DALL E?

[dusky linden]

You sexy fucks... how did you come up with a meta search engine that is lightyears ahead on bings copilot?

I like I like

Now if you wouldnt force me to login in and make an account every 3-4 prompts I'd like even more

Who can I thank for perplexity?

Hey ehmm.... how about facial recognition? I have a friend being bullied on social media with secret photos and videos taken of him and spread secretly in groups we cant find. Do you guys know of any software/ai equivalent that could help index/ scrape snapchat, instagram, twitter, telegram etc.? Could even perplexity do such a thing? I'll post this both in general and ideas!

[minor shard]

lol I do not think this is a thing.

Could probably somehow scrape twitter/instagram but not possible with perplexity

[dusky linden]

All i knwo is that yandex is currently the only good facial recognition service/ search engine..... BUT only works for the Russian internet and social media

I have tried the Apify console with their different agents. But none work for presumably well and definitely not for private accounts.

[old blade]

anyone has aan idea how to trigger the new Reasoning in pro search 100% ? i feel like its kinda random

[hasty sandal]

In web app when creating a new thread, open the focus setting. After "Web, Academic, Math, Writing, Video, Social" is Reasoning BETA.

I got a bad first impression of Reasoning beta. Compare The Unicode character U+237C, named "Right Angle with Downwards Zigzag Arrow" with this reasoning thread, which hallucinated: https://www.perplexity.ai/search/what-are-the-oldest-significan-xLHSwkeDQWScqYbGeuQFFg.

[minor shard]

It really depends on the question you ask it. If it needs to pull a lot of information and then create something like an analysis of that information it’s likely it will trigger.

[old blade]

nahh this is wrong. This is the "Normal" O1-mini reasoning. Perpelxity added a Reasoning for web searches. They had it already 2 weeks ago than removed it and today a better versen rereleased.

[old blade]

Yea more likely to get it thats true. But still sometimes i need a specific analyses and he doesnt do it and sometimes fore less specific he does. It really feels random. The Old Reasoning model was wayyyy more often while this new one is much more "thinking" unfortunetly it also halucinates more often than the old one.

[minor shard]

Yeah I agree. I’m not entirely sure why they wanted to make it “automatic” instead of letting users decide that but I’m assuming it’s just to avoid abuse.

[dense jolt]

Did anyone find their threads missing today? All I'm seeing now is 'Get Started' button, when I had a list of saved threads for reference later.

[fervent needle]

+1 all collections missing and perplexity is very laggy

[kindred folio]

why all this info?

also ios app searching is still cut off when it’s searching a source

[trim orchid]

Looks like it's fixed now. I could see threads inside collections again

BTW why the "Reasoning" feature is only in desktop app but not in web app?

[sleek vortex]

desktop means web

there is no desktop app yet…lol

mac app isn’t out yet

[rustic pasture]

https://www.theverge.com/2024/10/15/24270774/new-york-times-cease-and-desist-letter-perplexity-ai-search-engine

[sturdy wind]

any chance i could get access to pplx mac app beta?

cc @fierce sonnet @signal hamlet

[full gyro]

A Pumpkin has appeared!

Time is up!

[minor shard]

Really interested to see where this eventually goes. If perplexity can’t, then why is google allowed to? Technically google has been doing this for way longer even before AI.

[tame current]

yea

was throwing me off

[tame current]

lmao

dude honestly

new york times is often full of bs in the most respectful way i can say this

im not paying money for their mediocre headlines

[dense jolt]

Agreed, I see them. While playing around with the collections when they were missing I placed one of them into 'Bookmarks', which momentarily appeared. Now, I can't seem to find bookmarks or a way to delete them. My bookmarks are blank at this point.

[full gyro]

A Pumpkin has appeared!

Time is up!

[versed plaza]

Is there anyone that understands llm nature? So I'm having issue where the model will not respect the token limit or the stop commands. Sometimes it will start its response by finishing my question in its own little way. Sometimes it will give me a response and then reply as me and keep going on forever. Other times it will give me a nice response and then start spouting gobbledygook and kanji. In almost all cases it just keeps going forever becoming less and less coherent before eventually cutting itself off mid sentence. I feel like it's something to do with prompt templating but I don't know anything about this. I don't think it has anything to do with the ram because I only experienced that issue in certain models.

[tawdry whale]

Our work has blocked all LLMs except Copilot. They haven't blocked the domains they've blocked any JS links which is an odd way to do it.

[rough robin]

hello, i love perplexity but for some reason i got rate limited on sonar large and huge. why is that? i am subscribed.

i can send 600 requests per model. the default model is unlimited.

[twin pier]

Hi! I want to know, what are the differences between ChatGPT & Perplexity AI?

[minor shard]

There is too much to explain here. You should play around with both and experience the difference. Will be far better than someone explaining it.

[twin pier]

Oh ok.

[crimson basin]

Did you ask perplexity? 🙂

[twin pier]

Yes?

But

[lime vector]

it cites sources and includes links and is much more accurate imo

im a student and theoretically it returns better averages on theoretical completion assignments

[alpine pike]

perplexity is basically a rag pipeline built around chat gpt

[tame current]

ai bad i dont like it

[tame current]

Fr

[serene sand]

https://x.com/perplexity_ai/status/1846287953599123757

Perplexity (@perplexity_ai) on X

Perplexity for Finance:

Real-time stock quotes. Historical earning reports. Industry peer comparisons. Detailed analysis of company financials. All with delightful UI.

Have fun researching the market!

Twitter · Oct 15, 2024, 8:32 PM

[dusky linden]

he's right it will disrupt everything. Now we have genz and gena who are passing classes without even knowing what they are doing. They solve everything with AI. These are our future engineers who will reach the workforce and be useless

Then we have those who arent tecchnically inclined for the jobs of the future and wont be able to do manual labor. People maybe 20 - 10 years close to retirement who cant be repurposed. Mayham will ensue.

[cursive jacinth]

It depends sir.

[dusky linden]

Dont come with ....oh it will create new jobs bs! And universal basic income will never happen!

Then we have immigrants from 3rd world nations pouring in to our country only to realize they cant be used anywhere and since they arent willing or cant go back go back are going to be at the brink of ruin and turn to illegal matters further destrying our society from within. Then our people will be forced to compete with them for jobs which will turn ugly

And since more unworthy college graduates will be getting their diplomas onyl with the help of AI, this will lower wages in their respective fields

The only good thing about AI will be AI girlfriends and AI porn I can script scene for scene.

P.S Is perplexity cucked and leftist?
I asked for the immigration percentage of specific counties and am getting deliberately incorrect answers. These values can be found in wikipedia fairly easily and even when directing the AI to wikipedia it doesnt want to give me the correct answers.

[junior quail]

Awesome

[tame current]

Good morning sir

[tame current]

Oh, I gotta go one for you

[tame current]

I read the whole rant you gave here and for you to come to this discord, which is for AI to complain about something you don't understand It doesn't make sense to me your statement which shocked me was the only good thing about AI will be AI girlfriends and AI porn I could script scene for scene And your other claim was AI has ruined education for children the statement you made about children is true but for you to say that all AI is horrible because of children abusing the system to pass is a ridiculous statement And how dare you talk trash about AI without any research any backup to your claims and you're lucky that mods don't ban you and I'm gonna prove you wrong. These are all the positive things that AI has done for us because you are such a small minded person that all you look at is something negative. Let's go through this together, but just for you I'm gonna put it in simpler words.

[barren heron]

Hi i think perplexity is better than chatgpt because it gives latest information and it is not perfect in math and chemistry

[tame current]

That’s great. I hope you enjoy it.

[barren heron]

Hi

Great to meet u

What do u think abt perplexity

...

[tame current]

Disaster Management and Social Good: AI has been employed by organizations like the United Nations for effective disaster management, highlighting its potential in aiding global humanitarian efforts and improving societal resilience

Environmental Sustainability: AI has contributed to environmental sustainability by optimizing energy consumption in data centers and improving fuel efficiency in transportation. These efforts help reduce the carbon footprint and promote sustainable practices

Business Efficiency and Decision-Making: AI has enhanced business efficiency by automating repetitive tasks, reducing human error, and providing reliable insights for quick decision-making. This has led to cost reductions and increased revenues for companies that have integrated AI into their operations

Scientific Progress and Innovation: AI has accelerated scientific progress by controlling hydrogen fusion, improving matrix manipulation efficiency, and generating new antibodies. These advancements demonstrate AI's potential to solve complex scientific problems and drive innovation .

[tranquil orchid]

Hey @tame current!

If you find the original message helpful, please consider reacting to it with the :star: emoji. If the post is appreciated by the community and receives 5 stars, it will go to the https://discord.com/channels/1047197230748151888/1082806833938436228 channel and the post author will get the <@&1082034222778302614> role on Perplexity.

[barren heron]

Disaster Management and Social Good: AI has been employed by organizations like the United Nations for effective disaster management, highlighting its potential in aiding global humanitarian efforts and improving societal resilience

Environmental Sustainability: AI has contributed to environmental sustainability by optimizing energy consumption in data centers and improving fuel efficiency in transportation. These efforts help reduce the carbon footprint and promote sustainable practices

Business Efficiency and Decision-Making: AI has enhanced business efficiency by automating repetitive tasks, reducing human error, and providing reliable insights for quick decision-making. This has led to cost reductions and increased revenues for companies that have integrated AI into their operations

Scientific Progress and Innovation: AI has accelerated scientific progress by controlling hydrogen fusion, improving matrix manipulation efficiency, and generating new antibodies. These advancements demonstrate AI's potential to solve complex scientific problems and drive innovation .

wat is this

[tame current]

Sorry that was supposed to be for the other person

I’ll talk to you right now hold up

[barren heron]

Okay

but i gotta go

Byeee

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@barren heron

[tame current]

What I think about perplexity it’s pretty good, but I normally use another service because of math and biology

[barren heron]

Can u name that ai

pls

[tame current]

Do you want me to give you the name of the AI I use?

[barren heron]

yes

...

[tame current]

I texted you on private

[barren heron]

Okat

tkx

[elder zenith]

anyone else found it crashes all the time lately

[twilit dawn]

who can make me a pro search i pay

[tame current]

you are being rude

im hurt

reported

[tame current]

Did you look at the gift I made?

[tame current]

wym

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@sharp steeple

[tame current]

Come on man I spent hard work on that

Look, I’m gonna put it again

[tame current]

lmao

[tame current]

took you that much work to keyframe it? lol

i appreciate it though xd

thanks king looks good

😂

[tame current]

You have a great day now I’m gonna go to the abyss. See you later alligator

[tame current]

have a good one

[twilit dawn]

who can make me a pro search i pay
$

[tame current]

me

i can do it for free lol

he just ran out of it probably

and he aint paying, probably just baiting

ill help anyway

[twilit dawn]

cause i only need 1 time

fr?

yoo

5 and resets every 6 hours

thats fantastic

tysm

[rugged lance]

How to set perplexity as search engines, what is the url like perplexity.ai/s=query?

[tame current]

Uhhhhhhhh

[tame current]

https://www.perplexity.ai/

[full gyro]

A Pumpkin has appeared!

Time is up!

[tame current]

There you go if they don’t work, I don’t know

[rugged lance]

thanks

[twilit dawn]

It looks like you might be feeling a bit overwhelmed or unsure. No worries! I'm here to help with whatever you need. Do you have a specific question, topic in mind, or something you'd like to discuss? Just let me know!

[outer jackal]

Is there an official perplexity extension for firefox?
Found this one when I searched on this discord.
https://github.com/pnd280/complexity

[tame current]

says the one who be writing an entire bible in general

im reporting to the mods

[vital fossil]

This place sometimes feels like a kindergarten instead of a Discord for AI.

[tame current]

perplexity is NOT ai

its just a place where u chat with multiple llm!!!@!@!@!

had to update firefox rq tho bc of the new exploit

[vital fossil]

I guess it's time to block somebody, cleans things up a bit.

[outer jackal]

Yeah, tried it a few minutes ago. Found the settings. It looks great. Like the customization.

[little cedar]

Doesn't perplexity have the same features as gpts that chatgpt provides? Is there AI specialized for certain fields? Can I create my own AI?

Also, when you send a picture and ask them to solve a math problem (especially a shape problem), they are often unable to solve it well even though it is an easy problem. Why is this like this? For equations, is the correct answer rate 100%? Is there no solution to this?

I would really appreciate your answers to these questions. (Please reply with a comment.)

Also, why does perplexity AI keep generating images strangely or unnaturally? Is this feature not perfected yet?

[craggy monolith]

What's the context window for Perplexity Pro if using Sonnet 3.5? Would it still be 200k? Edit: nvm, found out it's 32k.

[little cedar]

There are AI models and AI image generation models, which is the best? I would like to be recommended.

Also, how do you use inference mode?

[royal cargo]

hey how can we get the hat we get for like 5 signups?

(please ping me if you reply)

[mossy jolt]

👀

[minor shard]

3D!

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@minor shard
@serene sand

[minor shard]

Is this in connection with Apple’s 3D AI model lol

[serene sand]

https://www.testingcatalog.com/redesigned-spaces-and-purchases-coming-soon-to-perplexity-users/

[tame current]

No, that’s crazy. Perplexity is a whole Amazon now.

[minor shard]

lol

Nice to see that they keep trying to clean up their UI

Still not sure why they haven't made it easier to change models/"spaces"

[pearl cipher]

and ebay, and alibaba, it just keeps going

[tame current]

I don’t want to be a hater, but I think perplexity should stick with it searching instead of becoming a whole shop

[minor shard]

I would agree. Not sure people are asking for spending money to be easier lol

Maybe it will be good for those people who are not good at cross-checking prices and making sure they get the best price for quality? idk. I am assuming this is to compete with Google's shopping tab so you can find everything in one place.

pplx really trying to go at google's throat while its down with the DOJ lol

[pearl cipher]

best to get them while they are down

[serene sand]

https://x.com/AravSrinivas/status/1843889051377774941

curiosity

Aravind Srinivas (@AravSrinivas) on X

Naming a meeting room after my favorite mathematician was a long standing dream come true. Ramanujan is an epitome of curiosity, the core idea perplexity stands for.

Twitter · Oct 9, 2024, 5:39 AM

[lament pond]

Hey folks. Reasoning mode for programming is really amazing. It seems it really thinks and never fails. Fantastic! 👏🏼👏🏼👏🏼👏🏼

[minor shard]

That is great to hear. I've had some really good success with it too.

[tame current]

https://tenor.com/view/arangutan-monkey-dancing-gif-15130385

[indigo valve]

I used vettedai for that for a while

[minor shard]

Have you had good results and accurate results with this? never seen it before

pretty cool that is also cross-checks social media sites like reddit

[indigo valve]

It's pretty good results. I use this or thegigabrain

@minor shard gigabrain does something similar where you can look through reddit threads for options

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@minor shard
@indigo valve
@serene sand

[minor shard]

very useful! ty!

[full gyro]

🎃 Halloween Pumpkin Collection Leaderboard 🎃

🥇 @minor shard (𝓙𝓪𝔀𝓷𝔃𝓮): 5 🎃

🥈 @last storm (kentavius): 3 🎃

🥉 @sleek vortex (sneakyf1shy): 2 🎃

4. @rich sphinx (WhatIsAModel): 2 🎃

5. @serene sand (mares): 2 🎃

6. @halcyon coral (IceLavaMan): 1 🎃

7. @native latch (cruel): 1 🎃

8. @brazen vine (Space_Army8485): 1 🎃

9. @wet quarry (CoreMT89): 1 🎃

10. @crisp zenith (ilikecheese): 1 🎃

Page 1/2

[minor shard]

Pretty wild

[serene sand]

*** can *** you *** feel *** it *** anon ***

[tame current]

They trying to get 𝓯𝓾𝓽𝓾𝓻𝓲𝓼𝓽𝓲𝓬

[pearl cipher]

damn

[tame current]

This chat is more dead than my sense humor

https://tenor.com/view/skull-issues-gif-13031152103567454559

[fiery sierra]

Hello :V

[fickle sorrel]

how to tell perplexity to search for more stuff for my questions? do i just copy paste the first prompt and send that?

i want it to find all links from metabolic syndrom to ibs from 100 sources

[indigo valve]

I believe 20 is the limit for pro. The more sources, the more system intensive it is

@fickle sorrel Why not convert it to a page?

[serene sand]

Perplexity Head of Product, Frank te Pas ☕

[empty lotus]

Heard this model from Nvidia is better than 4o and Sonnet 3.5
https://huggingface.co/nvidia/Llama-3.1-Nemotron-70B-Instruct-HF

[indigo valve]

I feel like at the end of the day, up-to-date information is key

[livid mantle]

one sentence

[nocturne oyster]

能问题吗

[valid iris]

https://www.perplexity.ai/search/for-each-of-the-dow-companies-bpvwO14gSayyU.xQJgUvzg#0

I did what the announcement said and asked perplexity on safari on Mac one of the suggested prompts to trigger reasoning mode, and this response has 81 sources, and I have another one with 33 sources

I tried asking it complicated questions about the ballot, but reasoning mode didn’t seem to trigger, and the total results were less than 15.

So did reasoning mode cause perplexity to gather 82 sources?

If so, I can’t wait for it to be on mobile

Does reasoning mode use either of the chat-GPT o1 models?

If not, what model is it using for reasoning mode if I have my default model set to sonar huge?

Also, when I was using perplexity to research the ballot, I stopped the response and rewrote it specifically asking it to trigger reasoning mode, and it did not do that. If I had selected reasoning mode as the focus, would it have gathered more sources than usual?

[halcyon coral]

Hey @valid iris!
Reasoning uses o1-mini 🙂

[tropic beacon]

what does that song called?

[foggy lily]

hi, is the only way to do images on the right of the screen the little box of options ?cause i keep getting told he cant do them and im telling him he can lol that we did it before , cant remember how I did the 1st time any help is appreciated

[wraith crow]

The Zap! - The Mouse And The Opera Singer

[rapid wigeon]

Perplexity API is quite slow, I am talking about a range of minutes for a response.

[tame current]

didnt meta create llama or am i just tweaking

[rapid wigeon]

API Slow, why? Apes together strong

Oooga oooga oooga

[halcyon coral]

Hey @rapid wigeon!
Please specify the corresponding model, the token speed in tokens/sec and the time to first token 🙏

[fickle sorrel]

how?

so far i find pages feature horrible

its nothing else than non-pages...

like u cant tell it to edit a part

adn the results are horrible... like manual esearch will be faster

[rapid wigeon]

Is that a benchmark I have to do manually? I am not streaming the responses best I can do is time to last token aka the response received time, such as when it was sent.

The model I am using is the llama-3.1-sonar-huge-128k-online model. I'll benchmark, but it's approx 1 - 2 minutes for a simple question

I quote often receive 524 as well.

[sweet jasper]

My perplexity is now in french

When it should respond in english

In web phone

[indigo valve]

It has to do with read tokens

[fickle sorrel]

can u do the pages for me 😄

[minor shard]

I wonder if PPLX plans to implement a form of ChatGPT's memory feature 😛

[supple pecan]

No joke, the answer is yes and I literally have the screenshots to prove it. It told me lol. In the same thread it told me it willingly lied to me (exact words) and that the model you pick for searching (chatGPT, claude, etc) doesn't matter because you'll always get a filtered Anthropic answer despite the facade you're using the other LLMs.

[indigo valve]

I haven't messed with pages enough lmao. Have you tried the chrome extension complexity?

[fickle sorrel]

hu hwhat that

is there anything else than perplexity wherfe i can let it go browse for x minutes?

perpl really fails at doing research for you

[slender bison]

ChatGPT with Wolfram Extension Thingy, but You can only do so much a Day before it not lettin Ya, cause You only get so Much Free GPT4 before it cuts You off

Also, Getting Perplexity or AI in General to do Math is… An Experience

[indigo valve]

You'll thank me later if you Google complexity extension

[indigo valve]

It's too resource exhaustive like I said if you wanted to continuously do it, especially with the amount of tokens it requires

[icy cloak]

https://www.perplexity.ai/hub/blog/introducing-internal-knowledge-search-and-spaces

[slender bison]

Very Cool

[manic kestrel]

in the pplx-ai web interface for pro users, which models can read and use attached files in their responses? sometimes the model's response pretends it has used the attached files, but very clearly have not read them, and sometimes the model tells me it can't read the attached files due to privacy reasons? it's a little bizarre at times

[kind timber]

I agree. Over the last 3 hours, I've been trying each model and non of them seem to be reading the file contents properly.

[sleek vortex]

what type of queries are you trying to do

[fickle sorrel]

i dont think queries is what matters. its that u cant edit specific parts of a box on "pages".

you can only edit one full box

and you dont know what will the result be.

[sleek vortex]

how is pages related to researching for longer periods

[fickle sorrel]

oh longer periods. i wanna use an llm to do medical research

like go thru all sites and make a list of known and possible causes of metabolic syndrome.

it's hard giving u explnation if im not doing it right now and i dont have pro

[keen seal]

I had a "collection" that I would use for writing papers. Now it's a space and it doesn't let me choose "writing" as a source? Only web?

[distant shadow]

Is anyone having issues uploading files to Spaces?

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@halcyon coral
@serene sand

[halcyon coral]

Hey @distant shadow!
There is currently a problem with file upload, a fix is already being worked on 🙏

[distant shadow]

Thanks so much for confirming 🙂 working on a new video about it atm as it's an awesome feature

[tranquil orchid]

Hey @distant shadow!

If you find the original message helpful, please consider reacting to it with the :star: emoji. If the post is appreciated by the community and receives 5 stars, it will go to the https://discord.com/channels/1047197230748151888/1082806833938436228 channel and the post author will get the <@&1082034222778302614> role on Perplexity.

[halcyon coral]

Hey @keen seal!
You can disable "Web" 🙂

[sleek vortex]

so...what happened to all the other focuses...

[halcyon coral]

You can use all the other focuses outside of Spaces

[sleek vortex]

I'm not inside a space

and the selector only shows "Web" and "Org Files"

possibly an oversight?

[halcyon coral]

Just to be sure: You can't see this?

[sleek vortex]

nope

[keen seal]

ahh, thanks! And we are phasing out the other sources (academic, social, etc)?

[lime vector]

will perplexity ever ipo

[serene sand]

+1 @signal hamlet

[keen seal]

I think it's a good idea to simplify the sources but would still love to have access to more sources in spaces (for example reasoning)

[regal loom]

Hii @everyone, I took Pro Monthly Subscription 5 hours back, but not satisfied with the results. Now, I wanna cancel my subscription and get refund. Can someone help me to cancel my subscription and get my money back.

[wheat tartan]

so, no more collections, just spaces now?

[sleek vortex]

1. most public servers don't let you just @everyone - not how discord works
2. go to settings > manage subscription > should be able to cancel there

[serene sand]

#📘│faq message please check Cancellation and Refund sections @regal loom

[wheat tartan]

YouTube: I started using Claude 70B recently and I gave it a youtube link and was able to ask queries about the video. It says that this does not work with Live streams. How is it actually working? And can this use case be expanded?

[sleek vortex]

how would it work with live streams?

it would have to download and transcribe in real time...?

since youtube doesn't generate a transcript or captions for live videos

[wheat tartan]

I fed the link after the stream ended

[minor shard]

It would still need a transcript/captions, did the video have one after the stream ended?

[wheat tartan]

No it did not, and on the one that it did work on, I didn't request transcripts, but it may have automatically done it

[serene sand]

https://x.com/aravsrinivas/status/1847030982211522852?s=46

Aravind Srinivas (@AravSrinivas) on X

The Answer Truck drove more than 1000 miles using FSD from California to Texas. And it's stopping by in Austin tomorrow for a Perplexity user meet-up. La Volta Pizza, Downtown Austin, 1 pm (Austin time) tomorrow. Food, merch, drinks, and check out the truck. Ty @Quadrant!

Twitter · Oct 17, 2024, 9:44 PM

[sleek vortex]

looks 😋

[minor shard]

is this ai

[versed plaza]

Do any of the devs use this discord? So I have come to realize that it is much easier and more coherent to have a conversation with photos in it if you upload to an external photo host and then just throw the link in rather than using that image as an eternal knowledge source, confusing any future queries within that conversation.

My proposal: if perplexity had a way to upload files to temporary web host such as postimage and then simply insert the URL reference within your message (the way some reddit apps do for example), then we could easily implement visual references into the conversation without fully derailing any future chat within that thread.

This seems like something that could be done through perplexity or the complexity extension, currently I have just been doing a lot of copying and pasting and shortcuts but realizing this process could be automated.

Currently the image upload feature only stands useful for certain situations. You could post a picture of a project and then continually reference that picture. But you could not use it to share ongoing photos of your progress within the project. Using URL links this is possible however. A way to utilize external temporary image host so that a link can be discussed rather than a raw image file would make perplexities ability to provide thought out analysis much more powerful and contextual to the flow of a longer conversation.

Absolutely but this problem has persist for over a year now. So kind of left to either build my own alternative to perplexity or throw out suggestions that well maybe people have been throwing out for months

[frosty vessel]

so, is spaces supposed to be a competitor to notebookLM?

[valid iris]

@halcyon coral I have a pro subscription, but no access to the pro channels

[hidden rapids]

Ok

Can I come?

Please

[tame socket]

where is o1?

There is a bug now, when returning from space to Home, it also becomes like this.

[whole fog]

How can I remove a file I uploaded to space? Didn’t find “delete” button or anything like that

[worn monolith]

I don't think there is one. I ran into that problem, too

[wide stratus]

LOL SpaceX is keeping the MOST interesting aspects secret... r u kidding me? 😩 😩 😩 https://www.perplexity.ai/search/how-much-computational-power-d-8F0FsqqvQCiq4tp25LnjgQ

[whole fog]

So it’s a new bug, looks like I have to create a new space before they fix this. I hope they will, I have used this space(or collection) for nearly half years.

[harsh verge]

#👑│pro-lounge message

[jolly pine]

any one see that the search speed is quite slow, im using pro version

[wraith crow]

you will have to leave the discord and rejoin with the discord pro link that can be found on your settings page on pplx

[rapid wigeon]

Rich Boi life

[rapid wigeon]

API is slow

[whole fog]

Thanks, but I found that it will downgrade the quality of the reply. So now I set to no use any resources and then attach files(not upload to space) to regain the original experience.

[weak pike]

Hi guys!!! Newbie here, so I apologize if asking this on the wrong thread l, but how do we collect these supposed Halloween treats!?!???? I’m just not really understanding it for some reason…. Thanks in advance!!!!

[serene sand]

#📰│server-news message

[onyx scaffold]

hey guys, I love the new voice input, I can talk and also can edit if needed, that’s what I was missing for a long time, thank you so much

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@weak pike

[weak pike]

LoLl!!!!:🎃 Well that was easy…,
So it just depends on “good timing” in #💬│general then, I guess!!???!!!!??? Answered my own question LoL!!! 😂 Thank You @full gyro Looking Forward to collecting more!!!🎃🎃🎃Stoked to Be Here!!!!

[plucky pebble]

Do perplexity have O1 model?

[woven cosmos]

Using complexity extension -> https://github.com/pnd280/complexity

[atomic stirrup]

sources for api?

[sleek vortex]

api is basically abandoned

[atomic stirrup]

yo wait ive dmed u before

[quick lodge]

only o1mini

[plush cairn]

Is there something wrong with my connection? Lately, pop-ups like this often appear, so I can't use Perplex. Any suggestions?

[mossy jolt]

nope, just whipped it up in blender

[valid iris]

Is that from mobile or desktop? I haven't seen that option on mobile.....

[subtle ridge]

Those who using pro plan, hows the search in file work? Is it as good as notebooklm?

[woven cosmos]

Look at my message just below

[woven cosmos]

The Complexity extension remove this problem.

[tired quartz]

Is the cooling down time set to 1min rather than 24 hrs now?

[dreamy mulch]

yo guys any problem with api?

[valid iris]

I assume this only works on desktop? So you download this GitHub file (and I don't know what happens next) and it gives perplexity access to just one more ai model?

[woven cosmos]

https://www.perplexity.ai/page/Complexity-Perplexitys-New-yl0q3mHYQz6RhRyuvjvN4w

Read this article to understand

[eternal kelp]

Is anyone using Perplexity for enterprise? I would like to launch it at my company and I am looking for best practices.

[cobalt monolith]

Is it possible in free version of perplexity to bias answers to be towards specific sources like lets say britannica over wikipedia

[eternal kelp]

I think you can specify that in the prompt