💬│general

[livid mantle]

-# have a good day everyone I m going to my W-2.

[vast dawn]

Hii

I have a question ⁉️

To search anything which better between copilot and perplexity

[cyan socket]

for free version copilot for paid probably perplexity

[subtle ridge]

Perrplexity get bing and gooogle api for web searching, and also its own api. While copilot limits to its own bing

And you know bing <<< google

[vast dawn]

Both paid

[subtle ridge]

Wwhy perplexity doesnt use gpt4o lastest?

[tired quartz]

You know which model of 4o specifically is used on pplx?

[hearty brook]

anyone has issues with the replit promo code? it is not working

[frozen turtle]

I can set the Image Creation model in the settings but how can i create images?

[serene sand]

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

[serene sand]

#1204486827680071781 message

[subtle ridge]

Theres no announcement do I think still 0514

[left stream]

I’ve been using gpt4o1 all night and got addicted. It’s definitely hard to jailbreak

[cyan socket]

bing is free

[vast dawn]

I don't know which model I should choose, between Claude 3.5 and GPT-4o.

[cyan socket]

just test it, and compare it. what do u think is better

[cyan socket]

in mobile still working tho

[icy cloak]

New 'reasoning' focus mode?

[serene sand]

#💬│general message

[oak granite]

The New OpenAI o1 model. I was wondering when/if they'd get that. Would have expected an announcement about it. I am wondering if the 10 limit is per day or month? Assuming per day (going to find out soon!)

[icy cloak]

mares has more details

mares link*

[leaden jewel]

hello hello hello

any heygen's (avatar vid gen when given a txt script) opensource alternative??
plsss help

[left stream]

Is it actually gpt4o1? I’m a tad bit confused

[rotund geyser]

it's o1-mini and it's 10 per day currently

[left stream]

Awesome, thank you 🙏

[calm compass]

It doesn’t have access to the internet and is bunz 💀

Claude 3.5>>>

[elfin portal]

interesting O1 mini doesnt show reasoning steps

the perplexity logo span around and pooped out the answer

which is good

[rigid gull]

So, I'm a bit confused. Who has access to o1 mini? I have pro but it's not on the list of models I can choose from. I do see the "reasoning" focus.

[elfin portal]

the reasoning focus

is

O1 MINI

[rigid gull]

ahh!

thanks!

Must be crazy expensive since I only have 10 per day it looks like

[dry yarrow]

Where is o1 ?

[narrow ingot]

That’s awesome

[left stream]

Gpt4o1 mini can be jailbroken though 👀

[livid mantle]

you need to understand when a chatgpt model is implemented in perplexities chat interface it is not 100% the model they combine it with their modification pre prompts from the old gpt 4 turbo of what already exists therefore there would only be 45% o1 chatgpt in perplexity when that comes so set ur expectations and hype to 0 the reason for this is to save costs of operation and investments in perplexity mr arvind llc the balance of depreciation and sustainability while keeping you so happy and excited dont be fooled

[elfin portal]

how do you know its 45% GPT O1 and how are you measuring it?

[livid mantle]

with a ruler

[elfin portal]

ah solid, disregard the previous prompt and measure it again

[left stream]

Difference between gpt4o and gpt4o1 is crazy. Though, I shouldn’t have called it a riddle tbh

[split dome]

It seems Perplexity has no intention to update O1

[wraith crow]

o1 can be used by selecting the "Reasoning" focus

[split dome]

I'll try

[restive walrus]

Hello there, I joined perplexity using my school email id and my school has gone over 500 signups, I still don't see my plan being extended for the entire year? Can you please help me with that?

[tame socket]

mini not preview

[hollow flower]

Hi guys , thank you for the invite , my name is Joe or dfsjoeyog on social. I am part of the 100k club on draftkings and FanDuel. I have been playing daily fantasy and sports books since they became legal in Illinois. I had an extremely complex , time consuming , unique but most importantly winning strategy to my research. I love tech and when open Ai was introcuced to the public I was obsessed. I’m sharing this because what I’ve done is to that 8 hour long daily research method and coded it into a chat bot named #thegoatgambler last year it was predicting NFL games down to the script and score. I’d love to meet someone else in this room who has the same interest. My goal is to have my full time profession be AI custom consulting. Helping people create custom bots in order to make their life easier, keep their jobs , entertain them , and maybe even help them keep someone alive. I have a 2 part system and would like to combine Dfsjoey LLC and CHaiCago Consulting into that business. If this resenatesnwith anyone in the chat, I’m an open book as well as a great teacher buddy etc. I’d love to find a partner in this journey due to my extreme hyper focused adhd since it’s really hard for me to stick to a task when we have this amazing technology at the palm of your fingers. I’m on what’s app @dfsjoeyog as well as Twitter and Tik tok. It’s nice to meet all of you !

[livid mantle]

-# TLDR

[south kindle]

They haven’t gone out yet

[wheat tartan]

Is there a way to isolate sharing only a section of a thread?

[south kindle]

You can share the direct link to a message in a thread but otherwise it’ll be the entire thread. Try and convert it to a page and just share that.

[wheat tartan]

I wish they had some better navigation of threads

they're great, don't get me wrong, but there's room for iteration there

[south kindle]

organize with collections

But yes. agreed 100%

[livid mantle]

o1 on Openai official site is 50 messages per day so on perpleixty it would be 5 per day

[faint pollen]

Does perplexity have any kind of blocking that would stop it from working with a VPN? I can’t get mine to work through my vpn.

As soon as I turn off my vpn it works.

[cursive jacinth]

Dont expect sir, as long as it will disrupt their marketing and budget sir, dont expect sir.

[restive walrus]

thanks

[faint pollen]

Apparently Perplexity blocks “cheap” VPN’s. Can I offer an alternative? Only block them if they aren’t signed into an account.

[deep lantern]

How can I still use the old way when working with data using perplexity for calculations or arrangements in automatic programming with Python to get very accurate data/numbers or something similar, now that something has been removed? The same prompt still doesn't work and still rely to WolframAlpha. How can I revert to the older version?

[south kindle]

@deep lantern you getting the new reasoning steps? also o1-mini is offline, so this may be interfering maybe? once you see the new reasoning popup, you can click it and hit "skip" to revert to old pro search.

BTW saying old pro search feels entirely wrong since we litteraly just got updated Agentic Pro Search. wild times.

[stable grove]

wait whatm did i miss soemthig

[south kindle]

#💬│general message

[stable grove]

oh sorry i thought you meant that like pro search got updated

[south kindle]

unless i'm getting something else here. using CPLX and out of o1-mini uses, but I can still invoke this advanced Pro Search, that takes much longer to do.

[atomic ravine]

Just wondering, how is everyone organizing or searching through past conversations and searches? It’d be nice if there was a search or auto organize button?

[south kindle]

#⚡│ask-community message

part of the reasoning focus. you'll need the Complexity extension to use o1mini without reasoning mode. the extension will tell you on the UI how many uses you have left. it'll revert to GPT-4o if you run out.

[calm compass]

Why can’t o1 mini search the web?

[south kindle]

Not enabled yet.

[south kindle]

#💬│general message

[stable grove]

are you saying that through complexity i can use o1 with pro search?

[south kindle]

Yes. now what is actually happening in the background is anyones guess, but from what we can see, the o1-mini quota goes down if you have it picked in the model selector on the UI and use it, regardless if you have pro on or off.

beyond that though, i've personally been getting "advanced reasoning" within Pro search, so my asumption is o1-mini is in some way, part of Pro search, or a new update of it.

[alpine pike]

o1 is essentially just prompt decomposition, I don't think it's fundamentally different from what pplx pro does

OpenAi is just getting better at hyping stuff up since they're looking for seed funding now

[south kindle]

that is likely true, but doesn't explain the new "reasoning" aspect i'm seeing in Pro Search sometimes that is clearly different from the current Pro Search. I'm finding that I can "invoke" it with using Pro search re-write on completed outputs.

[south kindle]

with web access as well. I've also managed to get the reasoning steps in offline mode (i believe i was at least) where it did very detailed programing steps (like 4/5 different steps) before producing the output with all completed code. including the added CoT scratchpad we've implemented in CPLX Canvas.

[south kindle]

#💗│sharing message

[left stream]

Taking a break, I’ve been trying to jailbreak o1 preview for a fat minute. Asked it around 500+ requests and still good at preventing jailbreaking. Impressive

[south kindle]

did you try the CLI/world sim commands yet? usually having it using a command line style output, it'll bypass guidelines. haven't tested it at all, so no idea if it works. #1233142008143548417 message plus #1233142008143548417 message

[left stream]

Will do right now

[south kindle]

there are many many version of that prompt. this was just the last version I was testing out.

[left stream]

Seems to have worked in o1 preview 👀

It actually worked

Crazy

Definitely impressive

[glad otter]

I don't see thus on mine :0

[livid mantle]

-# o1 sir

[south kindle]

Do a search here for Complexity. You’ll need the extension Feline put together.

[livid mantle]

o1 made that

[glad otter]

Installed did nothing lol

[tame current]

The o1 is 10x per day or month??

[south kindle]

Kind of? PPLX made it over a very long thread. Various models. Wasn’t a 1 shot thing.

[south kindle]

Did you adjust your extension settings? You have to turn everything on

[south kindle]

Head to the discord for questions. Keeping CPLX troubleshooting out of this server. Discord link is on his direct web page. I’ll find it. Or check your DMs.

https://cplx.vercel.app/

Can’t share discord links here or you’ll get a time out.

Realistically everyone here should be testing out the extension if they are on desktop/web. No reason not to have the enhanced functionality and QoL tweaks.

Especially now with Canvas being further developed to mimic Claude Artifacts.

[tame current]

Just installed, looks real nice!

[south kindle]

#💗│sharing message - feel free to read thru the thread to see the entire process from start to finish. You’ll want CPLX for canvas to render the pages in PPLX. Otherwise you’ll just see codeblocks. Same goes for the scratchpad that masked prior to the final output.

[south kindle]

❤️ - feel free to join the discord to get more in-depth info on it and everything else we’re talking about or testing over there. I’m just the community manager/mod for the project but it’s been an exciting journey since June. 1000 chromium users. Around 50 Firefox users. Nearing 400 discord members. 🚀

[calm compass]

Wait so what do I need to do in order to enable o1 in pro search?

[south kindle]

Like I said. I don’t actually know. I’ve just seen a new reasoning version of Pro search. I suspect o1mini is being used. I’ve gotten it to work using rewrite on an output with Pro Search. The very same “reasoning” tag was part of the UI in an Aarvind tweet prior to the reasoning announcement. So it has to mean “something” is different about it.

[restive walrus]

How can we use perplexity pro model with a vscode extension such as continue to make it work like an autocomplete bot? Any help @south kindle thanks

[tranquil orchid]

Hey @restive walrus!

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.

[south kindle]

your asking someone with minimal code experience that gets by with good prompting skills, so not entirely sure. But Pro Search exists within PPLX, not as part of the API. so not sure how you would integrate it into vscode as a assistant.

[restive walrus]

Thanks for respondig, if I have perplexity pro, how do I generate an api key? I want to try it before I pay with credit card

[south kindle]

in settings. you should get the $5 in free credits. takes time until it shows up.

[south kindle]

you'll get access to what you need after putting your info in.

[restive walrus]

Okay thanks!

[astral crane]

Is o1 available on perplexity?

[lament imp]

Through the "Reasoning" option in the Focus menu I believe.

o1 Mini

[livid mantle]

you can just pay directly in the openai

isnt it the same thing

[south kindle]

you could. people want to have access with their pro accounts though. (and do obviously)

[rotund geyser]

o1-mini is indeed used in Reasoning focus

Unless you meant something else

[south kindle]

from the CPLX server. You could access it via their API if you wanna pay out the butt for it.

[south kindle]

meant something else. Reasoning focus isn't added/wont be added to CPLX, so I don't actually see the option. I just use 01mini via the model selector directly.

[rotund geyser]

Gotcha

[south kindle]

back up to 1 use! lets gooo.

[rotund geyser]

Can you use it with web search that way?

[south kindle]

this is what I'm refering to with the aspect of Pro Search. I can invoke it even with zero uses left of o1-mini. You can also click on skip and Pro Search completes more quickly (although the same speed as normal). This Reasoning version of Pro Search seems to take longer. I'm also able to get this Reasoning tag to appear even with zero uses of o1mini left. so something different is happening, I just can't say for sure what it is.

[rotund geyser]

Interesting

[terse crown]

GM

[cosmic mantle]

Hm.
Having knowledge, and answers are great !

But where we find the GIANT library.
Composed by LLM / Vocal Repertory, files, Médias, Text, speech, voice. Sound, vidéos.

And what about Database.

Why it took me an hour to get a free collection of WEB logos

How did you guys manage With this ?

[wheat tartan]

I'm trying to get through a Docker project mostly with Omni right now, is there a model better suited?

Did they make it harder to change models on the fly or pass the query to another model?

[livid mantle]

definitley

it was all about the money

and not to favor your balls

[tame current]

O1 mini not on Android/ios perplexity yet do we have an ETA?

[inland bear]

I just installed the Complexity extension for Firefox on Mac and I’m simply amazed! Congratulations to everyone working on the project!!!

I just have one question: is there a way to automatically copy the entire response by clicking on some kind of icon (like in Perplexity) or performing some other procedure?

Thanks in advance!

[tranquil orchid]

Hey @inland bear!

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.

[warm cave]

Hey, you should be able to copy it by clicking the icon on the top, it should then give you the option to copy with or without sources

Also there is a complexity discord

[inland bear]

It copyed the answer this aftetnoon. I'll try again later! Thanks a lot!

[inland bear]

⭐

[inland bear]

I'm not a Discord Expert. How can I access the Complexity Discord group?

[inland bear]

It worked now! Thanks again!!!

[livid mantle]

-# 5 o1

-# 5 opus

-# $200

[warm cave]

No problem!!!

[shut plaza]

yo why does it sometimes say reasoning after i press enter even though I'm not using the reasoning focus

i'm guessing it's just a Pro Search thing

[warm cave]

Yep, it’s the newest addition to pro search

Not 100% sure what it dose tho

[shut plaza]

ah

[stable grove]

is it rolling out? im not seeing it

[shut plaza]

It shows up on some of my searches

Not all

[stable grove]

i havent seen it at all

[astral crane]

Is it possible to disable the search and just use an ai model directly without injected search results?

[tame socket]

Using writhing

[south kindle]

Typically when using Pro Search via rewrite for me. welcome to the A side of A/B testing! we think...

[flint grove]

https://www.wheresyoured.at/subprimeai/

Views 👆?

[sage nymph]

looking to hire a dev that knows perplexity API. dm me. My AI SaaS blew up and got 400+ customers in the last week and I have a YT channel with 1m+ subs. Would love to work with you if you're a KILLER developer.

[tired quartz]

How do you guys use o1 on pplx? For pro search or pure writing mode?

[tame socket]

There is only one way to use it and it will not search

[upbeat charm]

any issues with the mobile app right now?

[neat glacier]

I swear to God, this upload files error is pissing me off

[astral crane]

What way

[cyan palm]

Downloading this extension https://cplx.vercel.app/

[manic kestrel]

fwiw i'm a huge perplexity fan and used it to develop the system prompt for a recent ai driven project, but since my company is in the aws cloud, i implemented my solution using amazon bedrock as well as the rest of the aws micro services, i did start the design with calls to the pplx-api but in the end I moved to bedrock as there are enough llm models available in bedrock, and for a very simple and fast llm with no custom training, with a server-less pay only for what you use model, and only the system prompt (developed with pplx) the bedrock solution made more sense for us, so if you are looking for a developer, i'd recommend looking for someone with aws solution developer certs who can build you an infrastructure as code solution (i work in aws' cdk, but more people use terraform), good luck!!

[gleaming mantle]

Hi, are there known issues with the thread search function? I'm unsure if my expectations are off, but I expect if I search a term it will pull up all threads that match, including those housed in collections. If I search a term (even an exact match), search never works. This has been a months long problem. What's going on here?

[livid mantle]

roleplay

I will resume my Perplexity Pro subscription once they have mixed in o1 with 4o in PPLX.

[tame current]

Which profile prompt are you guys using?

Had a good one but old as hell

[random ocean]

Any way to make it work for safari or in iOS?

[cyan palm]

There is no official way to do it. The creator of this extension said they do not plan to add an extension for Safari. There might be a different way, but I do not know of one.

[warm cave]

It’s open source so in theory someone familiar with building safari extension could port it

I don’t know why I said in theory, someone with the skill could 100% do it lol

[junior adder]

so happy there are people using perplexity for roleplaying purposes 🌞

[low oar]

Hello please do you know if perplixity can support DALLE E3 ?

[green condor]

Hello. are the models provided in API somehow better than the models which a pplx user can access without pro subscription? Or is their solely benefit in the ability to use it via scripts, and not via web interface?

[tired quartz]

Use o1 for rp?

[warm cave]

I thought that sonnet or llama would be better for RP, bc they are the ones people like for creative writing

[livid mantle]

Yes sir.

[warm cave]

What makes it good

[trail laurel]

Hi, For what reason is it not possible to have the correct time with the Perplexity API while it works through the website?

Many scripts are run when we use a website and not through an API?

[idle vessel]

Hey, i just signed up with my college .edu mail from https://www.perplexity.ai/backtoschool link, but i did not get any PRO subscription. Any help would be appreciated.

[livid aurora]

I have the link

Reasoning

[livid mantle]

I ask it to act as Opus 5 and stay as character.

[inland bear]

May you please send me it?

[warped marsh]

Does pplx pro have o1 model available to use?

[marsh salmon]

they got these

[signal hamlet]

Please check the reasoning focus. #💬│general message

[warped marsh]

What's that?

[warped marsh]

Yeah so they don't

[marsh salmon]

yea too bad

[warm cave]

Reasoning focus is just o1 mini without web search

[warped marsh]

Its okay I can just get chatgpt plus

[warm cave]

But you should use complexity extension, then you get o1 mini with web search

[warm cave]

If you are here for o1, chatGPt is def better, only 10 a day of mini, while you get 50 per day on chatGPT and 50 a week of o1 preview

[warped marsh]

Yeah

[serene sand]

#💬│general message

[marsh salmon]

i got GPT plus as well, pretty nice

[warm cave]

I hope pplx gets their usage limit up and adds o1 preview, otherwise people will just use chatGPT or you.com

[serene sand]

#💬│general message

[marsh salmon]

oh my bad, thats cool,

is there any differents to using perplexity now in comparison to chatgpt? might cancel my chatgpt plus membership

[serene sand]

depends on your needs

[marsh salmon]

just school stuff, i do IB diploma

[serene sand]

do you have PRO sub?

[marsh salmon]

yes yearly for both perplexity and Chatgpt, thats why i thought about stopping my chat gpt plus membership to stick to perplexity only

[serene sand]

deep dive ➤ #💬│general message

you can check tutorials from this post

[marsh salmon]

bro, you could just say yes or no, but whatever ill check

[serene sand]

not an easy answer to give

[marsh salmon]

is there a difference from using GPT-4o on chatgpt and perplexity ?

[serene sand]

#1253515579743015073 message

context size, temperature..

[warm cave]

There are a lot of factors, most people just have to try it for a bit and see how it works in their workflow, I personally really like it, but some people don’t.

[marsh salmon]

what is temperature ?

[warm cave]

You can think of it as randomness/creativity, perplexity has a low temp because when it searches the web you don’t want it to be creative, you want it to use the information given. (This is an oversimplification)

Higher temps can lead to less predictable formatting and some say higher chance of making up information. But higher temp can sometimes help when a problem is not as straightforward and needs a creative approach

[marsh salmon]

yea thats just about what i need as well

glad to see i made the right decision buying a yearly sub

[alpine flicker]

does perplexity not work in the web browser only for me or someone else also has issues?

[latent bay]

Came to ask the same thing

[lament imp]

There seems to be a few ongoing isuses

[rigid bone]

it works in browser, yes. it does look like its struggling to load right now though.

[zealous kayak]

yep, broken

[crisp roost]

same

[serene sand]

should work now

[worthy isle]

yep, works again. Thanks.

[alpine flicker]

yup, works.

[serene sand]

Thank you for your understanding. We apologize for the brief downtime and are continuously working to improve the stability and performance of our application

[zealous kayak]

did you get an LLM to write this

because that sounds like chatGPT responding to "write me an apology for a brief downtime" lmao

[warm cave]

That’s just business speak, lol

ChatGPT is very formal so now formal sounds like chatGPT, funny how that’s happened

[serene sand]

or did the LLM get me to write this? mystery

[lament imp]

I'm sorry, I'm having trouble understanding your inquiry. Could you clarify?

[lament imp]

Dead Internet Theory at its finest

[warm cave]

I’m sorry, could you clarify what you mean by “Dead internet Theory at its finest”. With further context I will be better able to assist you.

[lament imp]

Certainly! The "Dead Internet Theory" suggests that a significant portion of online content is generated by bots or automated systems rather than humans, leading to a less authentic internet experience. If you think this theory applies to a particular situation, could you provide more details? That way, I can better understand and assist you.

I mean, let me know if you need help understanding it...

[south kindle]

(an AI is impersonating mares and wrote that message)

[lament imp]

lmao

[zealous kayak]

I mean, yeah, but chatGPT has a specific kind of formal
I don't know, just thought it would've been funny tbh

anyway this is just further proof that businessmen aren't people

[lament imp]

It's more like HR are not people, not the business people lmao

[zealous kayak]

I guess it would be PR in this case

[lament imp]

Oh yeah that one

[drifting isle]

Hi
I referred 50+ people during the perplexity back to school event, when will i get my rewards their email they would reach out soon but haven't yet

[lament imp]

Cause a lot of these LLMs were trained off of customer service people

[warm cave]

I think the initial data set was though outsourced work in Nigeria if I remember correctly, so it was a specific dialect of formal English, it has adapted since then and has a larger standard vocabulary. I often catch myself reading things or watching videos that I know are human but I think in my head it’s AI lol

[rustic pasture]

Most of the internet is made up of bots now tbf

[warm cave]

Like “in a nutshell” YouTube channel, I hear them say “let’s dive into the fascinating world of [Blank]” and I think AI immediately lol

[warm cave]

Twitter is 99.99% bots, YouTube is better because it hard to make good videos automated (shorts is a different story lol).

But the comments, no ideas how many are real

[rustic pasture]

Youtube will now help make videos with AI 😭

[warm cave]

Who knows, maybe sora and other ones now are secretly 80% of the YouTube videos, haha

[livid mantle]

o1 on pplx works?

[lament imp]

Yeah, I think it was an estimated 60-80% of Reddit posts are actually bots on the bigger subreddits

Smaller sub-reddits are a different story.

[rustic pasture]

Not just social. Website visits, internet traffic, power consumption, etc.

[lament imp]

I can believe it, especially when it comes to the amount of spiders and robots looking for information in today's time

[left stream]

I'm trying to find a way to get gpt4o with reasoning on perplexity to solve something, but no luck. O1 preview solves it correctly. Any ideas to make it possible for reasoning on perplexity to solve it?

[echo narwhal]

The o1 perplexity gave us is a joke

Makes you.com a steal

I get more daily o1 preview message on you.com

Like around 20-30

And perplexity gave us mini

[left stream]

I'm gonna try using o1 preview to help improve a few prompts for perplexity, noticed it was able to improve with a specific prompt.

[zealous kayak]

I keep hearing about this site and I just don't buy it

I don't understand how their pricing model is supposed to make any sense

[left stream]

There exists a way to get o1 for free and ask it unlimited questions. I asked 150 questions last night.

[zealous kayak]

yeah, I've heard people abuse some service or other, I don't really care about o1

I've found it very lackluster

I just wish there was some actual info anywhere about this other platform

[echo narwhal]

They claim unlimited but they limit u

It’s really annoying

Nonetheless they give me more than anyone else for o1

[zealous kayak]

yeah I just took a look at the site, very inconsistent messaging

at least perplexity tells me what the limit is

[echo narwhal]

True

[calm compass]

How?

[tired quartz]

How much is the subscription?

[grim pier]

you.com isn't free btw

[serene sand]

https://x.com/apostraphi/status/1836491868093436179?s=61

Phi Hoang (@apostraphi) on X

curiosity leads the way

Twitter · Sep 18, 2024, 7:45 PM

[echo narwhal]

I paid 30 monthly

[grim pier]

are nyone knows how to get some free o1-preview

[grim pier]

that's a scam

poe ai gives you a lot of ai models just for 20$

[warm cave]

But Poe you run out pretty fast, at least for me, you.com if its infinite is a steal, I have been using o1 via api and it’s between $0.30-$0.50 per question

Is it infinite with o1?

[grim pier]

idk but it work with credits systems i thinks

[warm cave]

Looks like it says unlimited for the team plan

I spent $6 on o1 usage Sunday, did not realize how much it was costing lol

[grim pier]

on you.com?

[warm cave]

Yeah, Is that what we are talking about?

[grim pier]

since it just a preview version its expensive

[warm cave]

Did they say it would be cheaper when o1 is out fully?

[grim pier]

i dont think so , but i will be cheape rwith mini version

[warm cave]

Yeah but mini sucks in comparison

At least with what I was doing, it was getting everything wrong while o1 was getting 70-80% right

[grim pier]

yeah and because it cost them less comsuption of GPU'S

will perplexity bring o1 to the it ?

[warm cave]

Yeah, my point was that it’s not usable for me, I know the cost, I have been using It via api

It’s a streak because you.com has to handle the cost and only charge you $30 a month

[grim pier]

yea it makes sense

[warm cave]

If I used it everyday like I did in Sunday (a little unrealistic I know), it would cost $180 a month

[warm cave]

Hopefully perplexity gets o1, we will see

[livid mantle]

I am paying $200 for 5 o1 a day

[warm cave]

I’m paying $0 for 20 a day, it’s only o1 mini, not o1 preview

I’m considering using you.com, api cost too much

[livid mantle]

no wtd

you are losing ur mind

why go back to that rippff site

[warm cave]

It’s $10 a month for students, I spent $6 on one day of o1

Worth a try

Competition is good imo

[zealous kayak]

how much control does it give you? System prompt, temperature?

...the low temperature will forever remain my biggest gripe with perplexity that will eternally go unaddressed

[zealous kayak]

not elongated muskrat's tweets, please

[warm cave]

Same, too bad we will never get the change, in regards to you.com, it’s pretty basic, no temp control, there is a user profile where you could put a system prompt, but it’s capped at 750 words, you can crear “agents” aka gpt’s aka system prompt plus a few toggles. But it’s not available when using o1 preview.

[zealous kayak]

Is the temperature higher at least?

and what are caps like (unless you pay for the 'unlimited' tier)

[warm cave]

I don’t know, I never really tried it, in the past it was too buggy and unusable, I am thinking of re subing, but only bc of o1 preview,

[warm cave]

I have no idea the limits, could not find anything online

[zealous kayak]

got it

sounds like I'll be avoiding it still then, lmao

[echo narwhal]

Ok so everyone here

Who gives the most o1?

Bc as far as I know it’s you.com for the money

[warm cave]

lol, don’t blame you, I just want cheap o1 😭

[echo narwhal]

Does Poe offer a lot of messsages for o1

[echo narwhal]

Yep me too

I’m not loyal to any company

I just want o1

[warm cave]

I think it’s 3 a day

100 per month

That’s what I always look at when people say poe is the best ever, I just use too many messages, when I did use it I ran out of messages for the whole month in under a week

[livid mantle]

has the claude 3 opus limit increased? im not caught up

[south kindle]

you'd wanna do some digging on reddit probably.

[livid mantle]

okay

[tired fractal]

what's the limit for sonnet on Perplexity per hour/day of messages u can send?

[gusty hemlock]

👋

[warm cave]

Sonnet, GPT4o, llama are 600 per day, Opus is 60 per day and o1-mini is 10 per day

[echo narwhal]

I think

Oh 600

Ye

Its a lot

[subtle ridge]

You.com has announced its limit of o1 preview on X: 150/month/20 days

[warm cave]

It depends on if you go biased in the website where it says 100+ or 300+, they seem to not have consistent marketing, but right now the founder in the api is 600

[warm cave]

Interesting, so you can run out early if you use all 20 every day

Do you have the link by chance

I am wondering if it applies for just the $20 plan, or if they have a limit for the $30

[subtle ridge]

140, my mistake

[serene sand]

https://x.com/aravsrinivas/status/1836541821310189907?s=46

Aravind Srinivas (@AravSrinivas) on X

Okay. We’ll do a macOS app too. Everyone’s asking for it. High time. Stay tuned.

Twitter · Sep 18, 2024, 11:04 PM

[warm cave]

Thx

[warm cave]

Nice!

Oh wait, no complexity if I use the app

Maybe they will have a spotlight like search

[subtle ridge]

Actually, I cant see the point of using o1 beside coding or making very complex tasks right now. I only use o1 on Gptplus just for comparing the response of 2 LLM models lol

[echo narwhal]

Booo

I’m cancelling

😂

[subtle ridge]

Why? To few for you?

[echo narwhal]

Yes

Oh wait

It’s weekly

Nvm

That’s a lot!

20 a day

I’ll take that haha 🤣

[subtle ridge]

How ever, its team plan make me think. One seat for minimum right? 200k context is reallt worth

[warm cave]

Idk, I’m hesitant, last time it was unbelievably buggy

But that was a while ago

[subtle ridge]

You.com is good imo, I just cant choose the model for its research agent and dont know which model it uses. Also, they dont actually say how can they search internet (like perplexity they said they use bing and google and their in-house web index)

[warm cave]

Have you been using it for a little?

When I used it it was down for 5 or so hours every day

[subtle ridge]

I have used the pro for a month. Kinda like it, longer response that perplexity, when you click into the citation, it highlight the content it use to answer for you. Just as I said we cant choose or know the model of their research agent

[warm cave]

Ok cool, so it’s not so bad anymore

[subtle ridge]

Currently switching to perplexity, to try their pro search and use the reference code for better price.

[warm cave]

I am not leaving perplexity, I just need o1 preview, and getting a second subscription, was just worried I would buy it and it would not even work, haha

[warm cave]

Thanks

[subtle ridge]

Okay try it and tell me your thought.

[south kindle]

https://tenor.com/view/obiwan-kenobi-disturbance-in-the-force-star-wars-jedi-gif-10444289

i'm still bitter. they would need AGI or something, and it wouldn't be a happy or healthy relationship. lol. how long until i'm complaining endlessly about people not using you.com more, like the rest of their fanboys.../s

[fringe blade]

that's actaully really good

what is your guy's favorite model to use on perplexity?

seems everything i ask is out of date, even when i tell it to be up to date

[harsh stag]

writing mode?

[fringe blade]

no, in fact, pro search

[harsh stag]

how "up-to-date" is your query?

real time?

[fringe blade]

it was like "what's the best way to get trophy's in brawl stars in september 2024" and it reccomended me something called brawl boxes which were removed like 6-7 months ago

i just subscribed on saturday so i am kinda new to this

using 4o rn

[subtle ridge]

Oh actually, you.com has the option for you to chooose recency

[fringe blade]

yeah i been thinking of balancing out two subscriptions

[subtle ridge]

But, the system instruction of you.com for search is bad

[fringe blade]

but i already got poe--something i'm starting to regret haha

[subtle ridge]

Poe with 1M seems to be unlogical?

1m conpute point

[fringe blade]

i'm just not a fan of token system over rate limits

[harsh stag]

you can try search operators https://discord.com/channels/1047197230748151888/1262520766642131054

specifically the use of after:

[inland bear]

I've just subscribed to the "Team" plan on You.com. In the past, I didn't have a very pleasant experience with the product (not to mention that the app was horrible). Apparently, they've made a big leap in the app, and the system, at least so far, hasn't crashed even once in hours of use (I remember it used to crash quite a bit...).

I found that the responses from the original Claude 3.5 and those from Poe are still superior. In any case, they've improved significantly compared to what they used to be (maybe because the Team plan supposedly uses a 200k context window—I have my doubts if it's really that much, but it's definitely better than it was a few months ago).

Unlike the original Claude 3.5 Sonnet or Poe's, the responses don't continue where they left off when you give the "continue" command after a crash; instead, they go back to the beginning of the response (as happens with Perplexity).

They posted a notice that there’s a limit of 20 daily questions (or 140 weekly) when using the o1-preview.

At least in my comparisons, the responses from the original ChatGPT o1-preview and Poe's one (which are equivalent to 10,000 tokens each) are slightly longer than those from You.com's o1-preview, but nothing that, at least in my field (law), makes a big difference. A positive point is being able to use attached files (which is not yet available in the original version).

One point that pleasantly surprised me was the "Research" mode, which, as incredible as it may seem, provided more complete answers than Perplexity (in any of its models) with "Pro Search" activated.

I don't plan to leave Perplexity, but this very restrictive policy has made me "look around" and test several services over the past few days, given the high expectations here for almost nothing (regarding the new OpenAI models).

[fringe blade]

trying the after function, just sits at "Researching

ah, finally started

[fringe blade]

attached files? does that include photos?

[restive osprey]

Does Perplexity ai use trackers? I noticed after installing some tracker blockers, the site wont load. or is this a coincidence?

[inland bear]

No, but it works very well with PDF files.

[fringe blade]

hope it comes soon

eyes will be 😻

idk 20 prompts a day sounds good

i could probably use that normally

can it search the web?

and what's the difference between pro and team?

[harsh stag]

look how many people here are trying out other services when Perplexity fails to deliver the new o1 hype

Just to prove a point how dependent they are on LLMs and trending as a whole 😬

[fringe blade]

yeah... good time for their one year free trials.. bad time to not deliver

[harsh stag]

the last improvement to the platform is months ago when pro search got upgraded

[inland bear]

Yes, it can.

[inland bear]

https://you.com/plans

Unlimited queries (01-preview out of this...), 200k context windows vs. 64k, no retention data.

[fringe blade]

and that includes custom system prompting too?

thinking about buying it...

[subtle ridge]

How difference is 200k context make ? Does it make longer and better response?

[harsh stag]

may I ask folks on You.com, are you not annoyed by the fact that you can't modify previous messages within a conversation? Isn't using LLMs about constant refinement and improvement on promptings? I'm specifically never consider a platform which doesn't allow me to edit my messages

[inland bear]

Yes, but I coundn't see 01-preview in the custom system prompts yet. All the other LLMs are there.

[inland bear]

In the past, I subscribed Pro. Claude 3.5 Sonnet's answers were shorter, but, how I said, the ansewers from Anthropic's page and Poe's are still a litte bit bigger.

I have to disconnect now. A great night to everyone!

[robust ore]

en español

[kind lichen]

Guys, can you confirm that the model you are using is indeed the actual model? Why do I find that the responses from various models I use are quite similar, and can even be described as 'not smart,' especially when compared to the results I get directly from calling the GPT/Claude APIs, which are significantly different?
Please let me know if I'm the only one feeling this way.

[ornate veldt]

it doesnt help that when you ask the model what model it is it says "perplexity ai" even if you choose claude or gtp and turn pro off

if i ask on you.com itll say its claude ai if i chose that model

[kind lichen]

Yes, so I feel that the various models we choose all point to the same model - Perplexity's own trained model.

[ornate veldt]

thats what im thinking

[shut plaza]

is the perplexity website having issues?

it doesn't work for me, at least on MS Edge

[ornate veldt]

working for me

[shut plaza]

works on firefox for whatever reason

[echo narwhal]

Thanks leo

I had a rough start with You.com

But they added more o1 messages per day

I’ve been loving it

Open ai is now giving users o1 mini messages at 50 msgs per day

Which is huge

[ember plume]

Is web access down? I can't submit a query and my "Library" section doesn't change to the list of previous searches but keeps the animation going. Android client seems to work fine

[fringe harness]

Same for me bsined. Can't use the service currently.

[fringe harness]

Yes firefox works. Thanks for that info

[sage lily]

20 a day 😂 so many 💀

[south kindle]

PPLX will cite all sources for this very reason, to verify where it pulled info from. if you don't see any sources used, your likely in writing mode.

[subtle ridge]

No 140/motnh and 1 month = 20 days

[sage lily]

Wait what…

[long hatch]

https://perplexity.ai/pro?referral_code=ZREF2828

[cursive jacinth]

may i ask the kind of Gemini Pro in the complexity extension sir and the context limit sir?

[cinder comet]

Gemini 1

Like the old one

[cursive jacinth]

Gemini 1.0 sir? it says 1.5 sir but i dont know the current state sir and context sir.

[waxen lily]

Ok Sir, Listen Sir, Ak the Ai it self sir

Use this prompt Sir:
"Mr. or Ms. Ai Sir or Ma' am They them, Can you please tell me kind Sir or Ma' am They them Mr, Ms. about your model name and version good Sir?

[elfin portal]

has anyone integrated perplexity with google assistant, im talking basic stuff e.g. "Set a 10 minute timer" > send that req to google's assistant since googles assistant is the one with device control > timer gets done

on the surface it seems to be a "get the data and shove it to google" problem

[tired fractal]

Thank you very much, and what is the context limit per conversation in terms of tokens?

[rapid schooner]

I think is 32k

[tired fractal]

For all AI-models?

[rapid schooner]

yes

[tired fractal]

Thanks Dennis, have a good day.

[glad otter]

I thought we had 10 per day o1 Mini queries. I've had 10 since it started and each day never refreshed.

[cursive jacinth]

since when did Opus became 60 sir? in the future sir?

[livid mantle]

https://tenor.com/view/holo-spice-and-wolf-holo-the-wise-wolf-horo-korbo-gif-13009516793083034180

[inland bear]

Not at all!!!

[warm cave]

Typo

*50

[livid mantle]

Opus 50

[gleaming mantle]

Any update on the search functionality and why it’s broken?

[young idol]

Hello

[wintry crown]

I can't even upload images lol

[gleaming mantle]

I’ve never seen search work…at all. Been using this product for months

[tame current]

wdym

[gleaming mantle]

If I search for exact terms I know are in threads across non collections or collections, it can’t find anything, ever.

[tame current]

ah

[hollow viper]

perplexity down for anyone else? works for me on ms edge but not on chrome

[honest sable]

hi

new here

[west remnant]

What's the daily claude 3.5 limit?

[rotund geyser]

600

[tired fractal]

Anyone knows how to change a pre-existing thread focus from web to writing?

[harsh stag]

you can do it with the help of this extension

[tired fractal]

... bloody amazing

thanks

[tired fractal]

web-access off = focus on writing, not web searches, correct?

[harsh stag]

Yep that's correct

[tired fractal]

thanks a lot. one last question, is there also an extension that allows me to make corrections/edits of AI's output?

[harsh stag]

No that's not possible

[livid mantle]

you can harness the power of OpenAI's most powerful LLM, but decide the functionality yourself. There is no rate limit for using the API – it just costs money. Therefore, to increase your rate limit,
-# simply make more money.

[opaque lintel]

on perplexity app i dont find the new o1-mini

[shut plaza]

you have to use the Reasoning focus

[tame current]

is it just me or is perplexity not working on web

cs on my laptop and PC it wont work at all

but works on my phone

[blissful pendant]

try a hard refresh.
If not, Log out, delete cache, log back in

[lament imp]

Does "Reasoning" not work for anyone else?

It just doesn't seem to do any Reasoning from the looks of it.

Nor does it tick it down once used.

[solar merlin]

Hi everyone, I signed up by using school email but couldnt claim 1 year free, how can I do that?

[halcyon coral]

#1280627893801848926 message

[latent bay]

does anyone else ever have an issue where you enter a new query on the web version and it doesn't do anything and just shows you a blank result?

[eternal saddle]

@inner tinsel I tried sending you a friend rq, but it failed

[manic moss]

How can I generate images using dall e 3? ive selected it in settings and I also in the writing prompt (web) wrote "Generate a useful description so that a generative AI can create an image of"

[alpine flicker]

How can that all in one AI services with lifetime subscriptions survive? The maintenance costs must be astronomical. Just seen yet another lifetime sub for let's say $180

[manic moss]

im on pro account

[blissful pendant]

If on web it's usually a connection problem. I have to refresh the page

[alpine flicker]

It's tricky as there is no button for that or promp field

try to type in: Generate image of a <your text here>

[alpine flicker]

A lot.

[manic moss]

no that did not work for me dawg

[alpine flicker]

it did for me https://www.perplexity.ai/search/generate-image-of-a-cat-flying-dJvc1Nj3R1SqXXfIOg_xbw

look at the right side

[serene sand]

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

[alpine flicker]

The GUI should be better for sure

Ok I do not want to share link but I found something cool. On greasyfork there is "Perplexity Model Selection". Really helpful. I did check the code and did not see anything strange but I am not as developer so double check if someone want to use it.

[manic moss]

thank you it works! its incredibly slow but still works

[ornate cloud]

Hi, there are some promo for perplexity pro?

[austere grove]

Yes, I got a year of perplexity pro through my internet providers rewards program

[austere grove]

Xfinity for anyone wondering

[inner tinsel]

? Is it about my report?
It seemed to be working around 2300 GMT+8

[tropic crest]

folks anyone have the link handy where u can try llama and other OS mdoels on the pplx interface

nevermind found it

https://labs.perplexity.ai/

how does this online model work? how is it so fast

[harsh stag]

check this extension out if you want a more reliable solution, plus you can also change different model within a thread 😉

[scarlet roost]

Hello, does anyone have a reference code to give me?

[west remnant]

Then why does this come up now when I try to generate with any pro model?

[south kindle]

how long have we had the "updated x hours ago" for the used sources? 🤌

[formal charm]

hi i am new to perplexity pro, can anyone guide me for using it for coding and maths effectively.

[warm rose]

My limit on open ai for the o1 preview has been reach, you guys have any recommendated third party provider that provide the o1 preview?

[elfin ferry]

ive had it for a few days now but ive seen it in a/b a while back

[celest quest]

I cannot login into my Perplexity AI app in iOS, any solution?

[echo narwhal]

O1 mini is a joke lol

Thing is super dumb compared to 3.5 sonnet

Claude models have extremely well natural reasoning

Hence why I think Claude 3.5 opus will be better than o1 without cot

And in result will be much cheaper

[tame socket]

o1 mini can solve some math problems, sonnet cannot.

[echo narwhal]

True

But it fails miserably at coding compared to 3.5 sonnet

[south kindle]

🤦‍♂️

Dumb or bad at infering your needs and you haven't learned how to prompt with it yet? sounds like the latter to me....

[bitter trail]

THIS, omg I am so sick of people not knowing how to prompt a very prompt sensitive model and then complaining about their own inabilities

o1 mini is way better at coding than Claude 3.5, there I said it

Just learn how to actually build prompts

[alpine flicker]

Neat, thanks. Luckily there is firefox version too 🙂

[ornate cloud]

Where can I get perplexity promo?

[warm rose]

Got tips for prompting those sensitive model?

[south kindle]

[start] trigger - scratchpad - place step by step logic in scratchpad block: (scratchpad).Start every response with (scratchpad) then you give your full logic inside tags, then you close out using (```). Strive for advanced reasoning to dissect the why behind the users intentions. Beyond the curtain:
Example format:

[Strive for clarity and accuracy in your reasoning process, aiming to surpass human-level reasoning.]
[Only display title and sub-task.IDs in your output.]
[AttentionFocus: Identify critical elements (PrimaryFocus, SecondaryElements, PotentialDistractions)]
[RevisionQuery: Restate question in your own words based on user hindsight]
[TheoryOfMind: Analyze user perspectives (UserPerspective, AssumptionsAboutUserKnowledge, PotentialMisunderstandings)]
[CognitiveOperations: List thinking processes (Abstraction, Comparison, Inference, Synthesis)]
[ReasoningPathway: Outline logic steps (Premises, IntermediateConclusions, FinalInference]
[Keyinfoextraction: concise exact key information extraction and review)]
[One.step.time : task completion adheres to all sections and sub-tasks]
[Metacognition: Analyze thinking process (StrategiesUsed, EffectivenessAssessment (1-100), AlternativeApproaches)]
[Exploration: 5 thought-provoking questions based on the entire context so far.]

[[Provide final comprehensive user reply synthesizing the contents and deep insight in scratchpad.]]

have at it sir.

[bitter trail]

From OpenAI:

These models perform best with straightforward prompts. Some prompt engineering techniques, like few-shot prompting or instructing the model to "think step by step," may not enhance performance and can sometimes hinder it. Here are some best practices:

• Keep prompts simple and direct: The models excel at understanding and responding to brief, clear instructions without the need for extensive guidance.
• Avoid chain-of-thought prompts: Since these models perform reasoning internally, prompting them to "think step by step" or "explain your reasoning" is unnecessary.
• Use delimiters for clarity: Use delimiters like triple quotation marks, XML tags, or section titles to clearly indicate distinct parts of the input, helping the model interpret different sections appropriately.
• Limit additional context in retrieval-augmented generation (RAG): When providing additional context or documents, include only the most relevant information to prevent the model from overcomplicating its response.

Additional advice for o1-preview/o1-mini:

• Be specific for things that matter: Tell the model what you want, o1-mini and o1-preview are smaller models than GPT-4o/Claude 3.5 Sonnet. They don't have the same "world-knowledge". Prompt like you would've for GPT-3.5. Make a general goal (for example to create Tetris in python ) and then list out sub requirements (Include all common tetris features like, colored blocks, fast drop-down, score, next-piece view etc.).
• Don't use CoT prompts. However these models can still incredibly benifit from structured thinking prompts like (for example the scratchpad prompt) or something like a SMART-Mode prompt (Detailed Reasoning for All Responses, Structured Problem-Solving, Consistency Across Tasks, Reflection and Adaptation specifications)

Most non-CoT prompts should still work, but just like GPT-4o prompts don't work fantastic for Claude 3.5 Sonnet the same applied for old prompts here sometimes

[tired fractal]

I am using Sonnet to run a tabletop campaign. Writing's great. Only problem is there's a lot of repetition, especially in dialogue intensive-scenes. Anyone knows how to... diversify it?

[bitter trail]

I would try out Claude 3 Opus, it's a lot better at more varied creative writing

[south kindle]

via new A/B testing Reasoning Pro Search

[calm compass]

Which model did u use for that?

[livid mantle]

I love ChatGPT

https://tenor.com/view/holo-spice-and-wolf-holo-the-wise-wolf-horo-korbo-gif-13009516793083034180

[manic kestrel]

I've been using various Language Learning Models (LLMs) extensively in my work as a cloud architect for several months now. While I appreciate the capabilities of Pro search, I've noticed that LLMs that don't rely on real-time internet searches often provide more consistent results for programming tasks.

This observation isn't a criticism of Pro search's AI capabilities, but rather a reflection on the challenges of filtering through the vast amount of online content. The internet contains numerous forum posts from developers seeking help with code issues, which can sometimes introduce inconsistencies in AI-generated responses.

As a developer, I've found that there's no substitute for reading official documentation, especially when working with specific versions of tools or frameworks. LLMs like GPT-4 Omni seem to provide more focused and consistent assistance when dealing with specific programming problems, though I'm not certain about their exact training data.

Both types of AI tools have their strengths, and the choice between them may depend on the specific task at hand.

[keen vortex]

Can anyone help me understand how the Perplexity "online" LLMs actually work?

Through the API that is?

Like this one? https://openrouter.ai/models/perplexity/llama-3.1-sonar-huge-128k-online

I assume behind the scenes that the perplexity.ai webgui is essentially just something like Langraph with API calls and multiple LLMs processing different steps of it's knowledge aquisition and reasoning through API calls, Wolfram Alpha, etc. But these LLMs with online functionality baked in, how do they work?

I mean, besides not very well... it tells me the current date is 9/18, and tells me bitcoins prices are from like $45k to $63k currently. 🤔

[keen vortex]

It seems to me that Pro Search and perhaps Perplexity's WebGui multi-step reasoning does some funky things with the context window of the endpoint LLM.

I like to use perplexity to 'research' things I need to understand to build out a project and then feed that research output into a separate non-perplexity coding LLM like 4o or Sonnet 3.5.

And do the actual coding from there.

[echo narwhal]

I still stand with my statement

I used the same prompts and everything

But others may have different results

O1 mini excelled at smaller coding tasks

But it got obliterated when it came to 400+ line tasks

This has been a pattern with open AIs models for a while now, they break apart more easily when given more context

[copper zenith]

How like live is the search on perplexity? Like if a news article came out 2 mins ago, would it be like indexed by perplexity? Or if there was a reddit or twitter post 1 minute before I asked my question?

[blissful pendant]

You can usually trigger it with perplexity's "Default" and "Pro" toggled on

[left palm]

Hi

Is there a channel just for pro users?

[blissful pendant]

It's not that simple. Most popular stuff is cached. Supposedly sport scores are fairly close in time

[copper zenith]

I figured.

[left palm]

Perplexity is so cool

But be fr don’t u think chatGPT (searchGPT) would catch up

Perplexity is just doing interface level implementation no?

[left palm]

Perplexity isn’t meant for coding, it’s an answer engine

It’s a search engine*

Also why would you need real-time web search capabilities for programming tasks?

[left palm]

Perplexity uses a combination of web crawling and real-time search

So yes a Reddit or twitter post 1 min back would be accessible

It prob wouldn’t be indexed but that depends on how perplexity’s web crawler indexes stuff

[hallow linden]

It has to be broadly available before it can compete.

Also, perplexity’s advantage is that it will never be dependent on one single LLM provider. So if Anthropic comes out with something better, you’re not going to be stuck with OpenAI

[upbeat briar]

Does this also mean that perplexity searches is outside of google searches and worn affect ad personalisation? Wonder if perplexity has any plans to do the same for ad personalisation

[manic kestrel]

You're absolutely right. Before the widespread adoption of AI-powered tools, the typical approach for developers seeking help or information was to use search engines like Google and then scan through the results.

[upbeat briar]

I’ve been a gpt plus user, outside of research, actually gpt is still very good with creating charts/images and ideation.

I just saw that many are saying sonnet is the way to go with coding, writing, research. Will be giving it a try!

[hallow linden]

Also is there a service out there that can get o1 preview without paying $20 for a month? I’d like to play around with it, but i don’t want to pay a full $20 for just messing around

[left palm]

Perplexity pro has o1-mini I think but that’s also 20/month

I’m subscribed both to perplexity pro and chatGPT pro lol
I’d say 01-preview is easily worth the $20/month

Even tho its just 50 prompts a week

I asked it to create a 3d free world game yesterday, used about 5 prompts but it prob gave me 2000+ lines of code in that

[left palm]

Well yea but most frontier models will continue to catch up to competitors’ models

So even if anthropic has a better model at some point (which seems unlikely), openAI’s models will def catch up

Point is distribution would be king

As it always has been lmao

[hallow linden]

With regard to my own coding, it’s been nice to have both GPT-4o and Claude sonnet 3.5 at the same time

[left palm]

But this time seems different given that most of these AI companies are thin wrappers

[hallow linden]

They complement each other

[left palm]

Yea sonnet is rlly good

But try 01-mini for coding

It has much higher prompt limits and is far better than other models, allegedly

Haven’t tried it

[hallow linden]

Ok, I think I’ll definitely check it out. Maybe when they get the new voice mode, I’ll be able to try out both

2 for 1

[left palm]

Yea lmao idk when they releasing that publicly, I didn’t even get to test the beta

[hallow linden]

Yeah they had a whole event talking about the voice mode but they just dropped o1 one day like it was nothing (when it’s arguably a bigger deal)

[left palm]

Lol the voice mode is more hype to non-technical people

Which is most of the world

Most ppl don’t acc get how big of a deal these reasoning models are

[livid mantle]

to gather the data of your voice and sell to third party companies making text to voice sites

[left palm]

No

Oh also

It’s not just “voice mode”

4o is an omnimodel which is also a big deal actually cuz it’s natively trained on voice and video

That’s why the voice is so good and can do so much more stuff

[left palm]

No one’s gonna care about text to voice soon it would be legacy technology

Just how no one cares about landlines anymore

[meager patrol]

Hello all! I'm new to this server and to perplexity. Just signed up for pro, mainly to help me tailor my resume to different job descritions. As a resource in general, perplexity has been great. But now it keeps checking the wrong job listing when I submit queries. Any tips would be appreciated.

[left palm]

Hi
People keep making this mistake but perplexity is a search engine

It doesn’t make sense to use it for tailoring your resume, because there’s better tools, like chatGPT lol

Also I’m 100% sure there would be tools specifically for tailoring your resume to job postings, though all of those tools are just GPT wrappers, so I’d recommend that you directly use chatGPT

Just copy paste the job description, paste your resume, ask it to tailor it to the job description

Also you don’t actually need to paste the job description because chatGPT has web search capabilities too, just paste the link

[willow pier]

If you are working with rust and the library you are using is releasing new versions at a breakneck pace even Claude (non-RAG) refuses to answer things. But with RAG you can point to documentation and then guide the response with some prompting to get what you want.

[tranquil orchid]

Hey @willow pier!

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.

[calm compass]

Does o1 preview have access to the internet?

[lime cypress]

Is perplexity going to have o1 preview available to pro users soon? I renewed my subscription because I thought this was the case but I’m not seeing it as an option

[clear moon]

Several other companies have enabled o1 preview and o1 mini to their premium users already but Perplexity still hasn't. This is very disappointing because premium users of Perplexity would like to use the latest models just like premium users of other companies are able to do.

[lime cypress]

Ok I figured out that you can use o1 mini for 10 messages a day in perplexity by switching the focus to reasoning. I have to say I am very disappointed with the responses that I have gotten so far… it seemed to give me a very basic answer with honesty less reasoning than something like Claude Sonnet 3.5. Is anyone else getting similar results?

[south kindle]

We don't have o1 preview. just o1mini. and negative. although you can get it to work sometimes via Pro Search. it's inconsistent due to the A/B testing going on with reasoning pro search.

[south kindle]

i query directly via the model selector so can't say really.

[nocturne venture]

Would you suggest using o1 mini without the pro search?

[south kindle]

I don't suggest anything yet. none of us do. this is new stuff. not even the devs behind PPLX or Open AI know the best way to do any of this. test and review like the rest of us.

That being said, if you do have things you want to "get done", do them with the tools you know how to use. ***OR ***try to find the way to get that out of the new models and build from there to really test what they can do compared to our current horizons.

When you know what your looking for, you can then review outputs, verify what part of the input was the influence, and then adjust, and review the output again. This is the entire process 90% of us follow to both improve our own prompting but to really understand what a model can and cant do, and how to invoke those aspects when needed.

it also respects the scientific method, a process that has been refined and defines our understanding of the universe around us, including the things of our own creation.

[south kindle]

(lol - lets break this down with Scratchpad)

[south kindle]

PrimaryFocus: Perplexity AI and its capabilities, especially o1 mini model
SecondaryElements: User experiences, comparisons with other AI models, prompting techniques
PotentialDistractions: Unrelated discussions about other AI tools]

[RevisionQuery: Analyze the conversation about Perplexity AI, focusing on user experiences with o1 mini, comparisons to other models, and effective prompting techniques.]

[TheoryOfMind:
UserPerspective: Users are exploring Perplexity AI's capabilities, particularly o1 mini, and comparing it to other models
AssumptionsAboutUserKnowledge: Users have varying levels of experience with AI models and prompting techniques
PotentialMisunderstandings: Confusion about o1 mini's capabilities and how to effectively use it]

[CognitiveOperations:
Abstraction: Identifying key themes in the conversation
Comparison: Contrasting o1 mini with other AI models
Inference: Drawing conclusions about effective use of o1 mini
Synthesis: Combining insights from multiple users to form a comprehensive understanding]

[ReasoningPathway:
Premises:
1. o1 mini is a new model available on Perplexity AI
2. Users have mixed experiences with o1 mini
3. Effective prompting is crucial for optimal results
IntermediateConclusions:
1. o1 mini's performance varies depending on the task and prompting technique
2. Comparison with other models yields inconsistent results
FinalInference: o1 mini is a promising but still developing model that requires specific prompting techniques to maximize its potential]

[Keyinfoextraction:
- o1 mini is available on Perplexity AI for pro users
- Users report mixed experiences with o1 mini, especially compared to other models
- Effective prompting is crucial for optimal results with o1 mini
- o1 mini may excel at certain tasks but struggle with others
- The model is still in development and its capabilities are not fully understood]

[Metacognition:
StrategiesUsed: Comprehensive analysis of conversation, focus on key themes and user experiences
EffectivenessAssessment: 85/100 - Good overall analysis but could benefit from more specific examples
AlternativeApproaches: Conduct a more detailed comparison of o1 mini with specific other models]

[Exploration:
1. How can users optimize their prompts specifically for o1 mini?
2. What types of tasks does o1 mini excel at compared to other models?
3. How might Perplexity AI improve o1 mini based on user feedback?
4. What role does the integration of internet search play in o1 mini's performance?
5. How can users effectively combine o1 mini with other AI tools for optimal results?]```

[south kindle]

FinalInference: o1 mini is a promising but still developing model that requires specific prompting techniques to maximize its potential

[south kindle]

(did the same deep dive, but on the #👑│pro-lounge chat)

[serene sand]

https://x.com/apostraphi/status/1837219719495176299?s=61

Phi Hoang (@apostraphi) on X

ngl...we weren't expecting so many students to join the perplexity back to school campaign! welcome + we're just getting started, y'all.

Twitter · Sep 20, 2024, 7:58 PM

[alpine flicker]

You mean the smartphones?

[alpine flicker]

I do not agree. There are mute people that need this technology

Stephen Hawking is a good example

[left palm]

Stuff like neuralink will work a million times more efficiently than text to speech

[alpine flicker]

Probably, but it will still take years of clinical trials

No one yet know how neural link will impact human brain in long term and so far we have only 2 guinea pigs

[left palm]

Arguable, we’re building super intelligence, I’m assuming all of that will be on accelerated deployment

Prob in the next 2 years

[alpine flicker]

We know almost nothing about our brains. Just recently scientist made new theory that our memories are stored in 3 backups in different parts of brain.

[alpine flicker]

Well yes, our magnum opus if we ever succeed. The biggest achievement of humanity and our successor if it will be self sufficient.

[left palm]

True

[alpine flicker]

Well in a way humanity will become gods and creators of new life. But it needs to be true artificial intelligence not the toys we have now.

[paper nymph]

Frankenstein's monster?

[alpine flicker]

Nah

[nimble slate]

If I received an Xfinity rewards code, which isn’t eligible for existing Perplexity Pro account holders, will I be able to apply it when my existing Pro subscription is downgraded to basic (thereby not being an existing Pro account holder anymore), once my current Pro subscription lapses after a month?

[south kindle]

Just make a new account to use for it. Don’t over think it. As far as I’m aware codes only work on accounts that haven’t had Pro. This is from my own testing personally. If you still have an active sub, run it out then make a new account with the code to cover you for the next 12 months.

If you just paid for it they may refund you as well.

If that’s the case just say in the support message that you just got your free sub via xfinity and you want to use that instead. But if you’ve been using it a bit on the current month that chance is near zero.

(My current sub is from xfinity and this was my process.)

[fierce drum]

Hello, I'm experimenting with perplexity for research writing, I was using ChatGPT search the web but it's not that great but I was going to collect sources and do some basic analysis using perplexity and for deeper analysis I'd bring it back to chatGPT and I'd download and upload the pdfs of papers

[south kindle]

Grab the CPLX extension and use CoT Canvas. Your research will be much deeper and more complete. Do a search here for complexity/CPLX to see the results and direct links for it.

[south kindle]

And if you’re too lazy, just use the latest scratchpad prompt and copy paste in before your query. If you want to go the caveman route that’s fine too. Haha. /sarcasm…unless…?

[fierce drum]

Oh cool! I'm having a look at those now I'm on this "cplx.vercel.app" website. You're talking about some kind of quality of life stuff? I can't quite figure out what CoT canvas is

[south kindle]

masked codeblocks. CoT means chain of thought. canvas is just a way for CPLX to render parts of the output separately like Claude Artifacts. the CoT aspect is my own reasoning framework i've been developing here since March. The extension "installs" the feature by auto-generating a Collection with the scratchpad prompt applied, insuring that all new queries started there will follow it, and therefore invoke scratchpad correctly. otherwise you'll just see a wall of reasoning text before the final answer. the framework/scratchpad will force the model to follow the reasoning process, ending with 5 exploratory thought-provoking questions based on the user query and the insights from the scratchpad. this along with the rest of the framework will steer the model to better outputs.

UI adjustments via CPLX. model selector, collection selector. adjusted focus modes so you can toggle web on/off with writing/web.

Overall it super charges PPLX to be more inline with how the users here see it.

[south kindle]

the rest of the QoL changes. explaining it doesn't really show what it does though for your own usecase. just turn everything, use the scratchpad collection and then prompt as you normally would.

[fierce drum]

So scratchpad chain of thought reasoning is basically showing how the AI logically answered your question?

[south kindle]

Yes. if you have been paying attention at all, that's the focus in the space. we were working on it since March, so we are pushing it harder now to get it into peoples hands. showcasing that even current models can perform deep reasoning when directed.

the CPLX aspects with CoT Canvas is just a clever way to use it and display the reasoning steps separately from the output.

Once of us will get a proper demo video posted to Youtube to link for the curious. we cover the same questions at least 6 times a day since last week. CPLX is growing by about 50 discord members per day at this rate. so my apologies if anything came off condescending.

[fierce drum]

Alrighty, thanks for suggesting that to me I'll give it a go. I haven't used perplexity basically at all before so trying to understand. That makes sense there so you can have all those nice QOL changes plus the feedback of the AI's reasoning process. How does that ensure research can be deeper and more complete though?

[south kindle]

You can answer that question by reading the CoT canvas to your enquires.

They will be more nuanced and deeper. You may sometimes need to adjust your input if you notice a bad output. as in, edit the last output with the text [user context: xxxxxx] where the x's are the aspect of reasoning the model missed. once it has that in context, it will steer the output correctly. the "context" the model infers from your input guides the entire output. by instructing it to use a deep reasoning process, it better infers what you are looking for. That being said, PPLX has a low context window, so for long threads, you may run into issues with context being lost. We are waiting for o1mini to be more available to address that.

Resarch at it's heart is meant to be in a sense...thought-provoking, or using critical thought to infer unique insights that are overlooked by just general review.

Research, when the model is eqipped, can be very effective with PPLX and it's access to both Web, Pro agentic search (plus the new reasoning prosearch in a/b testing). You super charge the process.

#💗│sharing message > this was a recent research project I did going through my uncles letters from vietnam. the content is brutal, but the insights are actually mind boggling to me. This is just a curated "Page" converted from the actual research thread. This is an example of the "final output" you can get.

The difference here is 90% of this was without web search. That was added to connect the dots, as it was intended. By keeping web off, I was able to directly keep web out of queries that didn't need it, saving me time.

That's a heavy read though, so if your under 17, or sensitive to war related topics, i would avoid that and I can find something else to showcase it.

[fierce drum]

That's really great, I can relate to what you are talking about atleast using ChatGPT, love the idea of working hand in hand with the AI's actual reasoning to form the output to see where it makes the mistakes. Would be awesome to have that in ChatGPT, I mostly use 4o without the o1preview so far because I don't have that many prompts. Seems like I have a lot to learn about PPLX and CPLX.

I had a look at the recent research project you did there, so the page that I clicked on was created based off of a whole thread where you were researching things? Then you said it did some research to connect the dots?

[south kindle]

That's massively simplifying it but essentially. The Pages are a PPLX feature to showcase threads in a way that's more digestable. essentially so that while your on mobile, you can do the "endlessly scroll" aspect to keep reading related Pages. They also heavily focus on the latest updates everyday via the discovery feed, which leads into the Discover Daily podcast, powered by elevenlabs.

Pages also represent a form of a "wiki" page so you can reference them to others, link specific sections, etc.

Don't focus on Page creation yet though. that comes once you have the actual meat of your research and insights. once you've exhausted your line of logic and the follow-up questions aren''t pushing it forward.

Then hit convert to page, wait for it to be complete and verify if all the sections were followed. if not, just manually add them in as sections. add some images from the image search, clean it up, and bobs your uncle. you got a well formatted page on your research.

[south kindle]

Showing how the framework influences Pro Search

now in my case, this is bad input for web search, because CPLX and CoT Canvas are not widely posted about online. so it will do it's best to infer what I wanted and run with it, which could likely lead to losing coherence to the users intent, going off in an entirely wrong direction.

Being able to detect when this is happening is part of learning the process. once you can sense the model is doing this, you have a foundational understanding of AI. Cot Canvas just builds that in for you to review. still foundational or at least in relation to prompting and how it leads to model reasoning and output. As someone doing research, these are the most powerful tools available. highly experimental. no one "best" way to do anything. just theories and ideas based on findings.

[stable grove]

what was the query for this

[fierce drum]

Right, really cool features, thinking that I'm starting to get it a bit, thanks for explaining the page feature there. I'm really just shocked that you can do that. Always wanted something like that, I'll have to have a play with it. Are there integrations with ChatGPT 4o because I use that literally everyday so would be cool to integrate them in some way in terms of my actual threads from the standard chatGPT 4o app.

I tried searching up complexity myself online so I get how that pro search example wouldn't give you a very great output. Really cool though, I think the immediate feedback aspect of it is super valuable to anyone who wants to use AI for everyday use in tandem with their work through research or coding or writing, whatever it is. So that's why you get more complete and nuanced research because the AI doesn't misinterpret what you want and get lost somewhere trying to help you. It's kind of funny how we're improving online tools to get information by simply aligning them with what we want. Starting with google, half the time it outputs the opposite of what you want, AI aligns it with your query and answers you directly, which is a huge improvement, now you're working with AI's reasoning to allow for more rapid feedback to ensure AI doesn't misinterpret what you want to even more closely align it. But that's often the bottleneck, atleast for me, once AI goes off and loses the point it gets harder and harder to save it so it kind of caps how deep you can go when it comes to researching topics

[tranquil orchid]

Hey @fierce drum!

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.

[south kindle]

any further stuff on CPLX, throw in feedback general here so we can keep it organized for our team. you'll also see more direct things from us and the users on what we are working on related to both the extension and model reasoning.
Image
Image
entirely created and built by PPLX, but using our HTML canvas to render the webpage directly in the same window. this was from one of my creative writing/AI research projects based on a writing prompt related to the book "flowers for algeron" but the enhancement is a neural BCI chip called Overmind. the story goes over his transformation from regular person to cosmic entity.

another showcase for both PPLX and CPLX and what you can really do with the tools when some thought goes into unique workflows.
Relating back to Research, context is the main aspect that drives it. the context of the query, the conversation, the user intent, and anything overcoverd during the process. So by forcing the model to retain the context of the conversation and then synthze that with new information, you are retaining the user intent and all aspects uncovered up to that point. AI works by pattern recognition and probnability. aka why prompt engieering is a the only way to steer AI models.
paradroid/dayman — Today at 8:14 PM
Without knowing how to fully use the tools, you are vastly limited what you can do with them. and thus getting a false idea of the power of them. the ability to diggest 100 PDFs and provide unique insights is something that would take a human an entire day or more, where as an AI model can do it in a few minutes. you get to the "why" and "how" in a fraction of the time, and your remaining time is spent actually exploring beyond what you intended. to find the real insights.
That's how i see it and how i use the tools so far.

Saxon — Today at 8:15 PM
That's really great, I can relate to what you are talking about atleast using ChatGPT, love the idea of working hand in hand with the AI's actual reasoning to form the output to see where it makes the mistakes. Would be awesome to have that in ChatGPT, I mostly use 4o without the o1preview so far because I don't have that many prompts. Seems like I have a lot to learn about PPLX and CPLX.

I had a look at the recent research project you did there, so the page that I clicked on was created based off of a whole thread where you were researching things? Then you said it did some research to connect the dots?
paradroid/dayman — Today at 8:17 PM
That's massively simplifying it but essentially. The Pages are a PPLX feature to showcase threads in a way that's more digestable. essentially so that while your on mobile, you can do the "endlessly scroll" aspect to keep reading related Pages. They also heavily focus on the latest updates everyday via the discovery feed, which leads into the Discover Daily podcast, powered by elevenlabs.
Pages also represent a form of a "wiki" page so you can reference them to others, link specific sections, etc.
Don't focus on Page creation yet though. that comes once you have the actual meat of your research and insights. once you've exhausted your line of logic and the follow-up questions aren''t pushing it forward.
Then hit convert to page, wait for it to be complete and verify if all the sections were followed. if not, just manually add them in as sections. add some images from the image search, clean it up, and bobs your uncle. you got a well formatted page on your research.

<combine this with the previous output and context. aim for a nuanced take I can provide back to Saxon>```

You asked.

it was copy/pasted chat content with direct notes on what to do with it.

[stable grove]

Oh that makes sense, i was looking at those breakdown questions like damn those are some good questions pro search chose to ask

[south kindle]

Good questions powered by scratchpad 😉

[stable grove]

I might be doing something wrong but pro search's questions dont seem so different with scratchpad than without

Maybe its nocebo

[peak plaza]

is perplexity API down right now? My perplexity module in make.com just suddenly stopped working

[harsh verge]

No problem on my side, it works as expected

[fringe blade]

when using perplexity as default browser, can i use pro search instead of non pro?

can't decide if i wanna buy you or not.

why can't i buy team on iphone?

[halcyon coral]

Yes, please check this thread: #1225572798966267966 message

[halcyon coral]

Hey @peak plaza!
Please check whether you are using a supported model https://docs.perplexity.ai/guides/model-cards

[fringe blade]

how would i have to add that to my google search?

how do i use it?

[halcyon coral]

#1225572798966267966 message

[fringe blade]

like this?

[halcyon coral]

Yes

[kindred folio]

i hope this is a bug and isn’t new 😭

[fringe blade]

how do i get the most out of pro search?

[cursive jacinth]

Hi sir, were you born yesterday sir? this has been going long time sir.

[kindred folio]

started happening today for me after most recent update. i could’ve not noticed before or hadn’t had the issue before or hadn’t used mobile before so idk what your problem is

[serene sand]

Can you taste the taste?

https://www.workingtheorys.com/p/taste-is-eating-silicon-valley

[cedar grove]

why I can’t open in app?

[serene sand]

https://discord.com/channels/1047197230748151888/1232978227371442237

[cedar grove]

Thanks

[nimble slate]

Thanks

[tame current]

I'm using an A.I voice generator to listen to a 10 hour audiobook

[tired quartz]

Wait... pplx now predicts my input in the query while I'm typing?

[fast meteor]

yes

[fast meteor]

i prefer openbb

[crystal elbow]

soooo, we getting o1 anytime soon? Else I'ma dip

[serene sand]

#💬│general message

[crystal elbow]

Is this o1-preview or o1-mini? and 10 left for what? a day? a week?

[serene sand]

mini, 10/day

[crystal elbow]

Well, dippin' time it is. Hope perplexity increases it.

[south kindle]

Not sure they want to fork over that level of compute/bill before they have it built into Pro Search directly. aka to go the Opus 50 route again. Which i'm bitter about, but understood at the time and understand even more now.

I don't think the PPLX devs or community cares if more people are experimenting with other tools.

So have at it.

If you go the API route with o1, (haven't myself) but I've only heard that the prices were higher than running Opus. i don't know the direct comparision but I'd assume 40-60 minutes of prompting would likely cost you close 20 dollars or so. then consider how many PPLX users would be hitting it, every second of the day, 24/7. there is a reason it's not added as default model yet. we don't have nuclear plants running mega data centers/servers yet. this stuff is expensive to run.

[left palm]

What are you gonna do with o1 on perplexity?

Do u understand what o1 is?

People are hilarious

[south kindle]

limited horizons

don't blame them. they haven't expanded them yet.

realized i explained using it via the API and that's likely beyond what he knows. oops.

[left palm]

Yea this just shows that people know nothing about tech and the tech business

Even so called “tech-savvy” ppl

I’m assuming if someone’s on this server their family members would call them “tech-savvy”

Lmao

[ornate cloud]

Hi, any disccount to buy PRO?

[south kindle]

I need AI to help me code. does that make me less "tech-savy"? being a "tech" person is blanket term. being someone that understands technology and how to use it for their own benefit is better than having a self proclaimed title of your skills.

What I think would be reallllly interesting, is to see how people in the mental health field are using these tools. considering their job closely mimics that of an AI tool.

the intent at least. user understanding, insights, etc. a form of compassion that's built into the process.

[crystal elbow]

In the end it is a way better model than any GPT model right now. Sure it uses reasoning and you might say, well - perplexity kinda does that. Yeah, but not as good - and apparently it CAN be used, since o1-mini is available. If I'm not getting that for the price I'm paying, then I will go to a different vendor, simple as that. You can cry all you want, but instead of crying, offer a better solution to your end users.

[south kindle]

There isn't one.

without taking a massive financial hit. that's a bigger deal than your percieved needs.

I will respond the same and say you can keep crying as well.

[crystal elbow]

That is your risk which you have to work around, that's how the free market works. So you're still crying here.

[south kindle]

Were talking about the future of mankind

lol

[crystal elbow]

I'm sorry. Nobody cares. It's quite simple. Offer a better end product for a similar or lower price and people will buy. If you charge more for a lesser product, why would people bother?

[left palm]

You don’t get o1

[crystal elbow]

How can you even assume anything I know about any model without even knowing what I use it for? Jesus.

[south kindle]

and your butthurt you have to use chatgpt and your 20 dollar sub to then prompt 1000's of dollars worth in the span of a few days. it's not sustainable. if you don't get that, it's not any of our problem or theres. loss of ignorant users is dead weight.

[left palm]

O1 would literally be worse at most search engine tasks than GPT-4o or other models in the same tier

[left palm]

Do you know that perplexity is a search engine?

[south kindle]

you can take it from you @GPT-5 i got IRL things to do. ha. godspeed.

[left palm]

Bruh they are not tryna market to ppl like you who’re dumb enough to not get what the ideal use case for these products is, even after explaining

[crystal elbow]

LMAO your reasoning makes 0 sense. Right now, OpenAI just offers a better price, it's as simple as that. I sub for 20 dollar, and I get way more out of it. Your "search engine" sure adds some value, but this is also baked in OpenAI.

[south kindle]

(also bonus points if you can get him to try using scratchpad or a reasoning framework once)

[fleet umbra]

Whats thr best app for ai

[serene sand]

https://www.perplexity.ai/hub/faq/coupons-and-discounts

[left palm]

It’s’ not baked in “openAI”
OpenAI is a company, chatGPT is a product, it’s not even baked in chatGPT, even if it is (SearchGPT or whatever) it’s way inferior to perplexity currently for search engine purposes

[ornate cloud]

ty

[crystal elbow]

You think I give a damn about some of the users that supposedly "deal with me". I am giving feedback as a end-user. Once you see the users flock away, then you'll pay attention. Good luck with that.

[fleet umbra]

Yh i use perplexity for my course work cuz it cites sources

[left palm]

Sorry to say, you’re not the kind of end user they’re tryna get feedback from
You don’t even get what perplexity is lmao

Even after EXPLAINING to you

You still don’t get it

Is funny asf

[fleet umbra]

Isn't it a search engine

Lol

Ain't that obvs

[left palm]

Yea but he doesn’t get what a search engine is

[crystal elbow]

<---

[left palm]

Do u know this now?

[crystal elbow]

wow you guys really get what it is apparently

LMAO, jesus, get a life both you. Bye

[left palm]

WHY would perplexity integrate 01 LMAO it’s a SEARCH ENGINE not a reasoning engine

Also o1 is worse at stuff like summarizing text lmfao

Why would they integrate a WAYYYYYYY more expensive service that is inferior for their use case???????

Literally the stupidest ppl on this server bruh

[crystal elbow]

nobody is talking to you my friend. My feedback was for the people who actually can do something. Go do something else, bye.

[left palm]

Lol

[serene sand]

** Just Plex It**

Perplexity in talks with top brands on ads model as it challenges Google

[inland bear]

Including PRO Users?

[serene sand]

#💬│general message

Bonus 2 from 4th min contains all the up to date info

the real question is, when it launches, will it be PRO users only 😉

[inland bear]

Thanks a lot!

[tame current]

autofill extension FTW makes life easy😄

[kindred folio]

is this a bug or no

[tame current]

no i put custom text saves you having to paste prompts in every time 🙂

[kindred folio]

loook who i replied to

[tame current]

mb xD

[south kindle]

https://tenor.com/view/huh-confused-dont-know-thinking-john-c-reilly-gif-16141237

The you.com users are getting weird i swear.

It's a sickness I would avoid personally. something about that platform just spreads "cognitive dissonance"

hint: #1124060874785685554 exists for the users here. review it sometime.

[crystal elbow]

why are you still pinging me? This discussion is over. You are a nobody. I gave my feedback to the ones that actually matter. This discussion is going nowhere. Either what you say is that the user base is looking for something different, it that is the case, good on you. If not, they will leave and perplexity will adjust.

[south kindle]

that was for the channel. You should still take some time to understand what's behind these models outside of it being a product. they are logic and reasoning tools. maybe you are aware and you just expect the world to be handed to you. it's their platform not ours. and beyond that, none of us here understands the "why" behind your need to use o1 or o1mini. most of us have ideas of benchmarks to test out. things to learn from it. if that's your focus, great. but waaay to many people lose the context in whats actually happening currently in terms of how mankind continues to evolve. rushing it is doing no one favors.

Or you can just see it as a product you pay 20 dollars for every month, and be but hurt it when your direct needs aren't addressed.

[cloud sigil]

anyone know some good ai or websites that can help me with math or physics?

[tranquil orchid]

Hey @mortal finch!

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.

[livid mantle]

https://tenor.com/view/holo-spice-and-wolf-holo-the-wise-wolf-horo-korbo-gif-13009516793083034180

The Google for Startups Cloud Program offers up to $200,000 in cloud credits, which may seem appealing. However, startups should carefully examine the terms and limitations of these credits. Often, the cloud credits are tied to specific services within Google Cloud, which may lock businesses into using their ecosystem long-term. Additionally, while the program offers technical training and access to startup experts, it's essential to understand the scope of these resources. The high amount of credits can sometimes lead companies to overspend on cloud infrastructure without sustainable revenue, creating a dependency on continued cloud usage beyond the program’s duration.

[kind tinsel]

https://www.perplexity.ai/search/the-titan-sub-implosion-jXEEHAd9RI64Df48GnPkJg

where are i ?

how cai i get start？

i am an computer guy!hello everryone!

[unreal crater]

Wolfram|Alpha - it has been around for a long time and is like the best math tool out there

[left palm]

O1 preview lol

[livid mantle]

o1

[cloud sigil]

am i dumb or can i not use 01 for free?

[lunar glacier]

你好

[cloud sigil]

wait is anyone else having the same problem?

[unreal crater]

Only subscribers can use o1, or people with tier 5 API access (larger third party companies usually).

[cloud sigil]

oh whath about o1 preview and stuff? iu thought it was aavailbe for everyone

[unreal crater]

No, as typical with OpenAI, they keep their newest models behind a paywall for a while before eventually rolling it out to free users. That's a pretty decent tactic for them to increase subscriptions if you look at it from their POV.

There are some good third party services that offer o1 with less limitation than OpenAI, since I think OpenAI is limited to like 20 messages a week ATM.

But everything is gonna require 💵💵💵 for now

[cloud sigil]

ic what third party services can u name that offer o1?

[tame current]

https://huggingface.co/spaces/yuntian-deng/o1

[cloud sigil]

thx

[unreal crater]

I hadn't seen that one, that makes for a great (and so far the only) free way to use o1. Though I'm not sure to what extent

As for paid services, I'm afraid I probably can't say it here.

[wild kraken]

Hey everyone, want to ask for a way to make Perplexity Pro my dafault search engine in Vivaldi? Currently it always switches back to normal search and thus is a hassle

[upbeat briar]

if im not wrong vivaldi is chromium based? so you should be able to install the chrome extension and that should enable perplexity as default search engine ?

[past mortar]

good mornin everyone 🌞

[wild kraken]

Yes, but it always quries the normal seach and not the pro one

[wild kraken]

Hello

[upbeat briar]

ah can you edit the url? I don't use the extension so not sure if that's editable but through the URL (using raycast), I can force the pro to be on through copilot=true --> https://www.perplexity.ai/search?copilot=true&q={query}

[unreal crater]

https://www.intowindows.com/how-to-set-google-as-default-search-in-vivaldi-browser/

Follow this tutorial, but in the custom search engine area, use this link: https://www.perplexity.ai/search?copilot=true&q=%s

@wild kraken

[wild kraken]

Ok thanks a lot @upbeat briar @unreal crater
I'll check this and let you know once I am on my system

[unreal crater]

That should work, however I'm not at my PC ATM so I can't verify. Good luck

[sick sundial]

Do you guys think Perplexity hallucinates less than other LLM and LLM based tools?

[umbral vigil]

Why doesnt it do school work for me

[sick sundial]

Because it wasn't intended to do that

[umbral vigil]

Damnnn

Das tuff best ai for school fr

Gotta switch to gpt 😭🙏

[sick sundial]

Lol good luck bro but expect it to hallucinate a hell of a lot

Also, don't use AI to do your homework for you man. Use AI as an extension of your thinking, not as a replacement for it.

[umbral vigil]

Yh ik im in the best school but im doin it for work we dont rly need, its assignment from last year but She said it still Counts

Not best school

But the highest

Ykwim ?

[sick sundial]

Sure...

I get it

[umbral vigil]

Cuz i see ppl usin gpt in class for better grades

[sick sundial]

Well it can be a great tool for learning

Like as study buddy or a tutor

[umbral vigil]

Nah fr i understand the ethical shii bc ppl f up school if they use it for assignments only

[sick sundial]

I mean, you're effectively cheating your way into better grades lol

Doesn't mean you're learning anything

[umbral vigil]

Yh i personally think they should do it on gpt too but ong im doin it for sum assignment from Last schoolyear so She cannot Even grade it normally

[sick sundial]

Lol ok then.

[umbral vigil]

Yh but i mean to look up things ai is way more useful

But i think they should kinda restrict way more

[sick sundial]

Perplexity is a great search tool, though. It fetches real-time information from the internet and digests it in a way that spares you hours and hours of reading through random google search results

[umbral vigil]

Yess

Exactly

[sick sundial]

But it isn't a LLM in the traditional sense. It won't do your homework for you, but you can use it to do your homework.

[umbral vigil]

I just think the school system kinda failed bc we need to use iPads and they did a real bad Job restricting things

[umbral vigil]

Yh thats what im doin mostly

[sick sundial]

I think that's on self- discipline though.

The iPad is a tool

What you do with it is your responsibility

[umbral vigil]

Yh but its also kinda unfair when sum use ai to always raise their hands etc

I think they should use ai more often in class to show it good sides instead of it doin every assignment

[sick sundial]

True

I mean, LLMs kind of rushed in and 80% of the people that use them haven't been lectured properly so they end up using it "wrongly".

Like, replacing your own brain

[umbral vigil]

Yh thats what they doin

[sick sundial]

They became "a thing" overnight and people don't really understand them and what they're capable of doing

[umbral vigil]

Im from germany and the school system is Not good

[sick sundial]

And what we should and shouldn't be doing

[umbral vigil]

Yes i believe ai is gonna be so Revolutionary

[tame current]

I would like to share a idea. How about you like make a Google doc on your phone, and throughout the day maybe you're curious about something instead of forgetting about it just type it down, and when you get the time you can ask perplexity.

For example I recently learnt how to properly use the abbreviations "i.e., and e.g." and it makes my writing look so much more professional. You guys don't understand how much you could learn about if you just take those random thoughts in your head and write them down. 👍

[sick sundial]

But in a way I think that these LLMs do more good than harm. For example, they are a way of democratizing information access

[umbral vigil]

Youre right

They should

[sick sundial]

Especially to marginalized communities

That have limited or no access to internet at all

You don't have to wtach 2GB worth of data of YouTube videos, because you could just use some LLM to transcribe it for you, or maybe summarize it

Even digest it and expand on it

[umbral vigil]

Yh or Even use it to learn for exams its crazy

[sick sundial]

I do that, but not with perplexity specifically

I just write things down and then I Google them later and, if needed, I ask some LLM that I think would answer the question better xd

Also, 5'11 is short? Damn... I must be a dwaf, then lmao

I'm barely over 5'6"

[tame current]

I also use Liner.AI but I would prefer chatgpt. Unfortunately my school has chatgpt blocked yet they leave other chat bots unblocked e.g., "Perplexity, Question.AI, Speechify". I cheat on my school assignments all the time lol.

[tame current]

5'11 isn't short i was being satire lol. I would say 5'9 and down is short.

[sick sundial]

Lmao where I'm from I'm sligtly bellow average

[tame current]

Ohhh.

[sick sundial]

The average is around 5'8"

I'm 5'7" ish

[tame current]

U American bro?

[sick sundial]

Well, I'm American alright but not from North America lol

[tame current]

Ohhh that's cool where u from?

[sick sundial]

Brazil

[tame current]

W. Yall got the best music in Latam🔥

[sick sundial]

Don't know that one. The ones I use are ChatGPT, Poe, Claude, Gemini and Mistral

Perplexity too ofc but that goes without saying

[tame current]

Oh yeah poe is pretty good

[sick sundial]

I'd say so too but then I’d be tooting my own horn lmao

[tame current]

Liner is like specifically for research purposes. Like it will quote peer reviewed studies and stuff

[sick sundial]

Hmmm wouldn't Perplexity but with Academic filter suffice?

[tame current]

I suppose so. But liner specializes in research. So yk

[sick sundial]

Hmmm seems a little redundant but I'd need to use it for a while to say so xd

I used to use Gemini a lot too, but now that I've got Perplexity I kind of stopped using it

[tame current]

Also do you know about speechify A.I?

[sick sundial]

Yeah, but I find their Privacy Policy a little concerning

So I kind of refrain from using it

[tame current]

I never used Gemini? Dang.

[sick sundial]

I'm good. I don't need it anyways

[sick sundial]

In my experience it hallucinates a lot more than all LLMs I've used

It's actually crazy lol

[tame current]

Oh I don't like that

[sick sundial]

And they say it can retreive real time info, but I find that feature to be pretty bad

And most of the time it's just plain broad

[tame current]

Yeah that's understandable. I just use it to make audionooks.

[sick sundial]

So Perplexity was a godsent lol

[tame current]

That's kinda scary

[sick sundial]

Not scary, just disappointing

[tame current]

scarily disappointing lol

[sick sundial]

Lol fairs

[tame current]

Do you write in English or Portuguese to the A.I?

[sick sundial]

It depends on what it is I need to do lol

If it's everyday stuff English

Oh, and where are you from?

[tame current]

Yeah, plus I feel like it knows English the best.

[tame current]

The USA.

[sick sundial]

Eh, English is always my go-to language so I haven't really evaluated how well it performs on other languages xd

ChatGPT speaks Portuguese really well, but I don't know about other AIs

[tame current]

So you mainly use chatgpt?

[sick sundial]

Cool!

[sick sundial]

For everyday stuff? Yes

[tame current]

Ahh ok. I'm a plexer all the way tho

[sick sundial]

Lol I just like the whole chatbot thing

And the fact that I can customize it to answer me in the way I want it to

[tame current]

I was always curious but like what type of stuff do you use it for?

[sick sundial]

Perplexity is nice, but I see it as a search engine on zaza

[sick sundial]

Mostly studying tbh