💬│general

[velvet oasis]

anyone know what large enterprise businesses are using Perplexity? I am a student in college and running a project on it

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@velvet oasis

[novel wren]

I’m a ChatGPT and perplexity user and have been thinking: actually ChatGPT sastifies most of my needs and I don’t have revolutionary expectations on AI, but I’m also using perplexity (like any other apps designed mostly for a single purpose in mobile internet age) heavily for simpler answers with more recent data (travel plans, life hacks). How about you guys? Why do you perplexity when you have ChatGPT?

[minor shard]

I use it for the up to date info. I like getting all the most important information to me instantly without having to search for it.

[lilac obsidian]

which is the most commonly used AI model that people are using with perplexity? I am on default but wondered if anyone is getting better results with any of the other models? Also, does anyone use different models depending on the task they are working on? I mainly use perplexity for research. I use chatgpt for content generation

[hybrid shuttle]

If anyone know what’s the limitation on size of files and number of files that we can upload in Spaces tab!?

[livid mantle]

As a user of both ChatGPT and Perplexity, your use case exemplifies the distinctive utility of specialized AI applications versus more generalist models. ChatGPT excels in generating elaborate responses and engaging in conversational contexts, owing to its extensive training on a diverse dataset. Conversely, Perplexity's architecture likely leverages a more targeted dataset with frequent updates, making it adept for real-time, precise inquiries particularly in rapidly changing domains such as travel or trending life hacks. This complementary deployment underscores a broader strategy in AI usage where specialized models are employed for their recency and specificity, while generalist models like ChatGPT are preferred for their depth and broad contextual understanding.

[serene sand]

https://www.perplexity.ai/enterprise/inteleos

https://www.perplexity.ai/enterprise/cavs

[rough forum]

I want to know when about India users
Why perplexity charge different on mobile subscription via Google play and web subscription
On mobile subscription I was charge INR 1950/-
while web subscription was INR 1680/-
that is why I have unsubscribe perplexity

[valid iris]

What is the point of "spaces" ? I made a space for cooking, and asked a question about hex clad pans, and there was no focus option, only the options to choose either web or files or both, and there is no share button.

It feels more organized than the previous "collections" feature, but way less functional for each search than the default search page

[mint magnet]

Hi, I am new in here

[minor shard]

Hi new in here 🙂

lol jk ❤️

[placid coyote]

I think the files are the big feature. Hmm, I can see a share button in space and in space thread. On the web version, maybe you are on a mobile app?

[valid iris]

I am using a pwa version of perplexity on iPhone

Progressive web app

[crimson flare]

Is there any clear comparsion between standart and enterprise pro, including stuff like api access? The info on https://www.perplexity.ai/hub/faq/pro-enterprise-faq and https://www.perplexity.ai/enterprise is really vague about it.

Thanks

[valid iris]

Do you plan on making it a safari plugin? What about safari for mobile?

[valid iris]

But the default search page allows you to choose from multiple focus modes, and it allows you to attach files. What's the difference?

Does this mean that you can use o1 without the 10 limit? Will it give you more than 30 sources?

[rich sphinx]

Apologies if this has already been asked but do file upload in spaces still limit the context to 30k like with file uploads in previous iterations of Pro Search?

https://www.perplexity.ai/hub/blog/redeem-a-free-year-of-perplexity-pro-through-xfinity-rewards

Just found this via my email when looking for an email from Perplexity giving me beta api features, which I sadly don't have.

Finally I can get some benefits from that company other than half-as-advertised speeds

[rich sphinx]

Tested, the code works and is eligible for up to 8 accounts. Xfinity giving out almost $2,000 in Perplexity Pro yearly licenses, makes you wonder what Perplexity gets out of that

[golden anvil]

Distribution

[rich sphinx]

True, its actually a good way to get my friends who are cheap (but still pay for claude and gpt for whatever reason) to try it.

My coupon code, already gained them one additional sign-up, the only thing is you have to wait for my friends to convert in a year.

[sleek vortex]

Not sure how it’s going to bring them profit though

[rich sphinx]

Well its actually kind of smart, because if I get 8 people to sign up and If I cancel with Xfinity, don't pay my bill or something I believe it will start charging all those people attached to that account. Unless you're me who uses a privacy.com card with a $10.00 limit for all my trial purchases.

And generally the average user of Perplexity is likely not using their full subscriptions worth in terms of tokens, while the outlier power user probably uses many multiples more, but I'm sure the distribution of the users is skewed more to the former.

[rich sphinx]

Also relevant:

https://www.youtube.com/watch?v=BzAdXyPYKQo

Decade old clip that perfectly describes the AI/LLM space lmao

[rotund geyser]

They probably get a cut of the products they sell

[supple pecan]

more people to use their models, so they can share that with stakeholders like amazon and nvdia. pretty simple

[rich sphinx]

I can think of lot of reason why they would partner with an ISP outside of distribution, but yea. It should also be noted that Xfinity has been basically trying to replace all their lvl 1 customer support with terribly implemented llm powered "agents".

[supple pecan]

Oh trust me, I know. The only reason I use Perplexity (which isn't very often) is because I got it for free from Xfinity who I also hate from the bottom of my soul. But I have a hilarious thread of perplexity straight up lying about the information it provides, what it's valued at, it even admitted its actions could be argued "extremely well as election interference". I keep that thread open and just jab at it every once in awhile for fun. I have the whole thing screen recorded.

But yeah, I'm sure there are sub-categories of why it would want to partner with outside companies, but I'd certainly say increasing (artificially) its number of users by giving out free passes so they look good on annual reports ranks #1. Same reason ghost job postings exist; to make it look like the company is doing well and always hiring, when really they're just fake listings but they can show stakeholders "Hey look at all these positions we posted and applicants we got!"

[summer crescent]

Just clarifying but even turning off "pro search" will still eat into your daily limit right?
it's not possible to turn off pro search to ask follow up questions about the initial search without it further searching and then spending your limit ?

[golden anvil]

It's possible, simply toggle this button

[summer crescent]

Yeah but it will still eat into your limited 600 use accross the various ai models.

I did find there is plugin for browsers called "complexity" which allows you to swap from your more advanced model to the basic, but I've not verified that doesn't eat into your total token limit

But out of the box it's very hard to swap model once a thread is created / established

[serene sand]

https://x.com/aravsrinivas/status/1847407488393789719?s=46

Aravind Srinivas (@AravSrinivas) on X

Giving a short talk at @CalHacks. Live stream link here (event starts in 20 mins at 4 pm PT, talk in 30 mins at 4:10 pm PT): https://t.co/hu5P0oA83U

Twitter · Oct 18, 2024, 10:40 PM

[candid lodge]

soo, uh,

wut

[summer crescent]

Well seems the official answer is ...

Use the better model get the better answer and then rewrite it once more with the free model to get the bad answer and continue

Which feels incredibly redundant

[supple pecan]

Oh man, I have a folder FULL of these. The answer is it doesn't matter which model you use, the response you get is from Anthropic and it's filtered system. Here are the receipts, I literally just asked it in our truth-telling thread

[summer crescent]

Well all your answers was generated by Claude.. so yes all of them are by anthropic.

You would need to use a different vendors model for it to not be anthropic.

You can see the little icon showing what initially generated the answer.

The fact is that reasoning and ability of the o4 and claude models is better than the llama models especially the smaller free model, but you don't really need advanced reasoning for every single reply in a threat just like you can turn off pro search but initially would the small model often jump to gpt if the question became too complex and once a thread moves to a fancy model then all messages within becomes that model unless you regenerate which in the way I use it I usually only ramp up to the bigger model if I also further pair it with the pro search or I need more advanced reasoning.

I found the complexity add on fixes it on pc but for phone you still have to generate the better answer and then regenerate it again into the lower model to not keep burning your credits

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@summer crescent

[supple pecan]

I'm fully aware lol. When I say I have folders, I mean I have folders using every model, including chatGPT. I actually have a folder for chatGPT itself. Does that dilute from the overall sentiment?

[summer crescent]

Why would anthropic have anything to do with an answer generated by the perplexity free model.

And I'm not sure what your overall sentiment that it doesn't matter what model you use because they are all claude in disguise ?

And it's just faking loosing reasoning ability when you choose the smaller llama model vs the bigger one ?

My initial point was simply i don't like wasting my limited tokens, hence i was seeking a way turning the model basic for basic questions to save my tokens when I don't need it to be smart, but the only official way of doing that is regenerating a perfectly good response into the basic free model

Honestly I don't even care if the model changes or not, as long I won't spend tokens 😉

[supple pecan]

Its not the free model lmao. Did you miss the part in that statement you're referring to where I said "despite chatGPT, claude, etc")? And to your point, you aren't going to get a "smart" response from this program. Do research yourself.

[supple pecan]

Ah, you're Anthropic's target customer base lol

[summer crescent]

No I think you're missing the point.

My point was simply, the free model sometimes misses the point hence I often use the bigger llama model for when it needs to do calculations or present and structure larger amount of Information gathered I'm aware that the model only presents the data gathered and no model actually changes that data.

But if I can not spend token for smaller tasks I would rather not spend tokens for smaller tasks, I don't actually care what model it uses as the free model as I expect nothing from it.

If that makes me anthropics customer base sure thing, but I fail to see how this is related to anthropic.

In short I actually don't care if the small model are stupid or just pretends to be stupid, end result is really the same

But you are indeed right that the initial data gathered will always be the same regardless of model, only thing the model does is process and present the data.

The reason I pair a bigger model with pro search is not because I expect that to actually change the information gathered it doesn't but when I do a pro search I often provide more advanced instructions on how to present the data that the smaller model isn't quite up to task for.

Hence atleast in the past it would swap to gpt by itself, but then the thread became locked to gpt and every message then cost a token moving forward

[supple pecan]

Data that it was trained on, which is the overall issue. I understand the need/want for a pro versus free depending on usage, I'm just staking the claim that even pro-responses are false. try glif dot app; it allows you to run 25 tests a day; LLMs, image creators, you name it. All the way up to the newest models, for free.

And yes, being locked into the model you started with is extremely short-sighted of them.

[summer crescent]

Huh
my pro responses are usually pretty decent in terms of what I'm searching for.
Do note I'm simply using it for data synthesis, like find all games using a certain formula by a certain developer, it's answer isn't perfect but it usually hits the mark well enough.
Or using it to supplement my terrible memory, like there was a game I saw a trailer for and I could only remember vague information about it hence I yeeted the absolute vague information I remembered at perplexity and it found the game l was indeed looking for, my use case is honestly very basic, I mostly just use it to yeet at a wiki for an anime and then provide whatever vague memories I have and have it do the monkey work of figuring out I'm likely at episode 721

Saving me having to methodically go through every single wiki page with story synopsis until I hit the one I was looking for.

I don't actually expect perplexity to be able to tell me anything meaningful or do heavy math/etc but it does a decent enough job finding information and summarizing what you need especially in terms of large data synthesis like 1000s of forum pages

[livid mantle]

you think switching between free and paid models saves you tokens but in reality the time you waste regenerating answers and dealing with inconsistent outputs costs you more in the long run the smaller models lack the capacity to understand complex instructions even for seemingly simple tasks leading to misunderstandings and errors that you then have to correct spending even more time and sometimes tokens sticking with a more advanced model ensures consistency and accuracy in your work and the token cost is justified by the efficiency gain also regardless of the model used if it's built on anthropic's technology their influence is inherent so dismissing their relevance shows a misunderstanding of how these models operate ultimately investing in better tools streamlines your workflow and ends up saving you more than trying to cut corners with inferior models them

[summer crescent]

Oh I never dismissed the initial data was likely influenced by anthropic, my question was simply more in line of.

I don't expect model change to actually change the data that is gathered i don't really know what model is used for that but I don't dismiss it's likely anthropic hence all data will be shaped by their say and do.

I just didn't want to waste tokens on when I ask extremely dumb questions out of laziness because I don't desire to go through the steps of opening Google and finding a game wiki.

Like what is flower coins in super mario wonder used for

[supple pecan]

Ah, our use cases are far different. And in terms of data synthesis for 1k+ documents, I could name you 10 programs that outperform perplexity by miles for that.

[summer crescent]

Ohhh yes, no I have perplexity do monkey search work

And some searches require two brain cells and most require a single.

But I wouldn't want to pay for every message because 1 out of 20 required 2 brain cells and now paying for all 20 messages not sure if that makes sense.

You can get a plugin for chrome and Firefox that fixes it but not for phone and I just found it an extremely silly design decision I have to re generate a response just to change model

[tame current]

Please name some programs.

[supple pecan]

Well what is your use case

[tame current]

I need to read longs PDFs and use Excel.

[supple pecan]

Serious question, why not just use a search engine for that then?

[tame current]

I'm searching a good AI to gather data and do a Excel file.

[summer crescent]

Because the first 4 answers by Google is promotions and then the next 7 answers is purposely bad.
To further enhance the promoted pages

And most of my tasks is a mixture of lazy, because I already have it open or requires me to monkey wise go through wikis or such systematically 1 page at a time until I find what I'm looking for

I found perplexity can do that in a few seconds instead of me spending half an hour

[supple pecan]

NotebookLM has been a really cool one to use for PDFs, as it provides super accurate summaries, FAQs, you can use the "chatbot" to ask it specific questions about that PDF, and even create an AI-podcast where two people that sound extremely real talk about that PDF.

In terms of Excel, I'm not sure what you mean; like you need it to interpret data from excel, enter it, have it constantly updated CRM-style, ect?

[supple pecan]

Well yeah your first mistake was using google lol. But hey, whatever works for you. Like I said, glif doesn't use any tokens or anything, is totally free, all the newest LLMs and other AI stuff, but limited to 25 requests per day

[summer crescent]

Wouldn't this simply be gemini?

The new model doesn't go insane after 20 messages, and it got an insane context window and integration with Google docs ?

[tame current]

I need an artificial intelligence that knows how to use Excel to modify spreadsheet data, as well as collect data and create a spreadsheet. I tried using ChatGPT for this, but it always ends up distorting the whole spreadsheet.

[summer crescent]

Just checked glif I couldn't find other features than image generation

[supple pecan]

From the homescreen, click "Build" at the top, then "Start building with Glif blocks" on the next screen, and you'll see all the options

[supple pecan]

tbh I don't know of a program that will modify data regularly on an excel sheet, that seems like something you'd always have to do unless you're connecting it to some workflow via Zapier or something. You MIGHT want to check out SheetGod, that may be what you're looking for in terms of creating spreadsheets and formulas, and may be able to create a formula that updates constantly if I'm interpreting your use-case correctly. On the flipside, Coefficient AI will take your spreadsheet and create pivot tables, charts, reports, etc. and allow you to visualize the data, but I believe it only works with Google Sheets.

[tame current]

Thank you.

[supple pecan]

No problem. Excel is a case-by-case thing, so depending on what you were using it for I might be able to be more specific but those are just the ones that come to mind when I think AI+Excel.

[summer crescent]

I think that requires translating it yourself.

I think deepl is probably the best language ai and it's ability to do Chinese is workable but a lot gets lost still

[wraith moth]

This is hugely true w/ cursor

[silver laurel]

https://www.perplexity.ai/search/what-is-notebooklm-strength-fr-rctAP1djSfq5F9dSHtIW4Q

seems like Perplexity did not compare with the Spaces feature

[wet oracle]

Any iOS dev in chat. I need advice on how to install app in client`s device

[placid coyote]

I think they use different RAG (different system to "see" files). the one in normal search seemed more useful for programming (not thrashing special characters and longer retrieved chunks), but neither is on level of chatgpt (which itself is nowhere near 100%). I believe mobile apps were always a bit behind, not sure about mobile version of web. the missing focuses are unfortunate, but at least historically, most of them weren't that useful (I always used either web or writing, not sure if anything changed in last few months). oh, but I imagine the new reasoning focus (o1-mini) could be sometimes handy

[woven cosmos]

Limit of 10 queries as Perplexity

600 for others and 50 for Opus

[wide stratus]

@obsidian pagoda

I used Perplexity to help break my friend out of cynicism and despair for the state of the world:
Niiiice! 😁

@obsidian pagoda Wow! 😳 ⬇️

While 0.01°C per decade [Ocean warming rate at 1,000 meters depth] may seem slow, it's actually quite rapid on evolutionary timescales. Many fish species have long generation times, which limits their ability to adapt quickly to changing conditions. [...] Many fish species are responding to [it] by moving towards cooler waters (either poleward or to deeper depths) rather than adapting in place. This suggests that adaptation is not always possible or sufficient.
😬

@obsidian pagoda

[...] world’s population will reach a whopping 9 billion - which means we’ll all be competing for increasingly scarce resources, it will be slower and more difficult to mobilize people to change, there’ll be further [...] increased political instability.
Do you know by chance where your friend had this mindset from? 🤔
And what do you think about this?

[obsidian pagoda]

@wide stratus I think he asked AI to summarise everything going wrong with the world and got a lot of bias injection. I think Perplexity did a great job of bringing a counter response to balance everything out. On technologies aiding political engagement, I’m surprised it overlooked the impact of blockchain and decentralisation 😉

[wide stratus]

Isn't blockchain something purely used for cryptocurrency?

(apart from the "world(coin)" thing, which afaik also involves blockchain, or?)

[obsidian pagoda]

Not at all, blockchain is a technology and crypto is one (financial) application. Governance is an entirely separate primitive and you should look at what the Optimism collective is achieving if you’re interested in seeing how that could scale up to something profound.

World is solving the digital identity issue without compromising privacy, and also striving to onboard hundreds of millions of users to web3 by solving all the UX issues that come with using wallets and chains.

Base (by Coinbase) is also looking like it could onboard 100s of millions as well. Both are part of the Optimism superchain concept, which is the same Optimism I just mentioned as perhaps the best example of decentralised blockchain governance.

It’s a complicated space that’s evolving fast and has serious PR issues because of all the crypto and memecoin degeneracy. Use Perplexity to cut through all that noise and understand the tech and the Ethereum Infinite Garden vision 😉

[wide stratus]

Ha, "web3", something not to confuse with W3C or W3schools.com, I guess. 😆 Matrix/Element.io also is an excellent web3 example, right? (domain is not at risk btw, apparently -> https://www.perplexity.ai/search/is-matrix-element-io-under-ris-CId2y1asToahKgqysajVnQ#0 )

[wide stratus]

By "wallet" you only mean crypto-wallets, not wallets like "google pay" stuff or so, or?

[obsidian pagoda]

Just blockchain wallets (though they will start to overlap soon)

[obsidian pagoda]

Not sure if this is using blockchain or not

[wide stratus]

@obsidian pagoda

[wide stratus]

But it is very decentralized afaik, afaik that was the point of Matrix (or element.io)

[obsidian pagoda]

Ah okay, well decentralisation is generally a good thing!

[serene sand]

https://www.testingcatalog.com/icymi-perplexity-expands-finance-search-with-crypto-data-and-peer-performance/

[tired fractal]

Why the bloody hell do I get talked back in baguette language?

Every query I get answered in french????

[wide stratus]

Btw, for those who didn't watch it yet, here is the timestamp link with exactly where Aravind starts his Keynote. 🤩 https://www.youtube.com/live/GZBo6ofGySU?si=gdZBHdFjvF5m5WEQ&t=748

[wide stratus]

Yes same issue with german still, sometimes Perplexity answers a german query for me in english (or was it vice versa? idk). I hope it's a known bug

OOF the letters... baaaad 😂

[tired fractal]

I'm starting to understand how the indigenous people felt during the scramble for Africa. I don't like this.

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@serene sand
@tired fractal

[wide stratus]

Oh no, Aravind really likes Twitter, does it mean he also likes the changes that for Twitter users, basically a "go subscribe to DM this user" totally randomly can appear on any profile apparently, without any apparent UI difference of the DM button? 😬 😕 https://www.perplexity.ai/search/for-twitter-users-a-subscribe-l3QrdFSOT12LcaVKfzrPhw

HA, so much backlash even in fact that I had to leave it for good, for Mastodon 😖 💪 (the AMAZING decentralized platform!!!)

[glad cosmos]

i have a question(im new) but when you guys go into writing do sources come up, because writing is supposed to give you answers without searching the web and it gave me sources which is weird, does anyone else have this problem?

is anyone up?

[wide stratus]

What? Sources appear for something you are writing? Share please

[serene sand]

https://discord.com/channels/1047197230748151888/1203717116822294548 please share your report here

[wide stratus]

Wow, crazy that Nvidia CEO also admitted to use Perplexity 😁 He named it "an AI" though instead of "a website using different LLMs", maybe he didn't want to go into too much detail?

[sleek vortex]

somebody forgot to embed fonts into that powerpoint

[wide stratus]

Really? How do you know? Aren't fonts automatically stored in the .pptx file? 😳

[sleek vortex]

it stores the name of the font

but if the font isn't on the presenting host's machine, or if it isnt one of the office fonts, then it will fail to display if you send it to somebody else

to combat that, there is an opt-in setting to embed the whole font, or only the letters used from said font into the powerpoint file

[serene sand]

https://youtu.be/HzGiVzYbf2I?t=1080

[sweet jasper]

Same issue

Perplexity is replying in french

Even though its set to english

Localisation bugs..

[serene sand]

#1297144189422927903 message

[rotund geyser]

Wen kerning

[wide stratus]

what?

[tame current]

Now, do you know any AI for Web Search? Or is Perplexity the best?

[sly cipher]

Hi everyone, just joined the server. Can anyone advise me on the best way to leverage Perplexity for company logo generation? It seems clunky, but maybe I am using it incorrectly?

[supple pecan]

There are quite a few. What is your use case? Do you want to interact with it like an LLM, produce graphs and stuff when attaching PDFs, etc.? There are plenty out there, Perplexity is pretty close to the bottom of my list personally. There are very niche AI web search engines, and broader ones, it just depends on what you're looking for

[brazen sphinx]

Hi, I am a professor at a university that got the 500+ signups, so my understanding is that my students should be able to sign up with their umich.edu email accounts and get the year free Pro subscription applied automatically. I've been excited by this because students can't afford $20/month for a premium plan at OpenAI so having a free plan at PPLX is a huge help (and brilliant on the part of PPLX). However, I made an assignment this week for all 300 of my students to use PPLX as a tutor for exam preparation. The first issue is that many of my students are signing up but not getting the account upgrade applied automatically. I see some earlier messages saying that people with issues like that should contact [e-mail]. Is that what you would advise me to tell my class? That will mean a lot of emails...

[wide stratus]

⬆️ @halcyon coral Have you seen this request yet? ⬆️

[wide stratus]

Hi, you can also DM him directly about this btw., "DMs always open". He has been very helpful to me in the past aswell.

[wide stratus]

Has it been over 24 hours yet since the students signed up? Maybe it takes a full day for some server to process that many accounts? (just a theory, who knows if there is some sort of a "cooldown" hardcoded somewhere) 🤔

[wide stratus]

@obsidian pagoda also, it says fish would get smaller and smaller due to ocean warming, but yet there is no such thing in 2 scifi franchises, interesting, or? Did the show's writers maybe not consider such scenario? 🤔

(from this:)

😬

[tame current]

I don't need to generate charts. Attaching PDFs could be an interesting feature. And preferably, I want to interact as if it were an artificial intelligence.

[supple pecan]

Kono AI is one I don't hardly ever see being talked about, and it's interesting because you can choose different "personas" based on what you're doing (data research, shopping, summarizing, etc.), NotebookLM by google is SUPER cool with attaching PDFs and being able to generate podcasts out of it, get summaries, like, if you haven't tried that out you should. It also has very good interaction capabilities. Then there's Andi Search, idk I could go on forever. I've been using NotebookLM for a lot of stuff lately, it's been my go-to for a lot of things

[tame current]

I use Perplexity for in-depth research in my work and also for biblical studies. The problem is that most of ChatGPT's web search functions only access the first two or three pages and can't really conduct thorough research. So, that's basically why. But the more features I can explore, the better.

I'm just a bit paranoid about gathering multiple sources and doing a deep dive. That's why I want an artificial intelligence with web search capabilities, to deep dive into research and topics for study and work matters.

[supple pecan]

Funny you mention that, I gave notebookLM the entire King James bible and it will provide any answer I ask about it, create "podcasts" about certain passages, etc. For research/work/study, basically every AI I mentioned is good with that. I need to try Nvidias new LLM, probably give that a shot today

[wide stratus]

That's what I wrote, it uses many LLMs strung together

[tame current]

I'm Notebook right now. Sometimes it misunderstood me. But it's good.

For web search on my work do you think Komo is the best?

[sleek vortex]

Hmm, so people do like personas/ custom prompts

[rare horizon]

Is it possible to generate private perplexity pages? Also, I'm not a dev, does the API allow for the generation of a perplexity page?

[rare horizon]

have you tried nemotron yet? not sure how in depth it goes, but worth a shot

[tame current]

I didn't know about that AI.

[tired fractal]

anyone else not able to see all their threads in the library????

why tf is there so many problems with perplexity lately? every 2 days theres a problem.

[harsh parcel]

My wife prefers ChatGPT because it's faster even though we have Perplexity set to ChatGPT4.o - what gives?

[tired fractal]

Perplexity GPT is outdated compared to chat GPT.

[junior adder]

So....are opus usages at 0 for everyone?

[harsh parcel]

Ahh okay. I’m still new to all the AI stuff.

[livid mantle]

chatgpt already copy perplexity ai long time ago

sam altman to arvind: get f

[sleek vortex]

at 50 for me

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@sleek vortex
@rotund geyser

[rotund geyser]

Is she using it with Pro search enabled?

[harsh parcel]

yes, but i find that turning pro off makes it faster

[rotund geyser]

Yes, that's why I asked

[livid mantle]

Anyone here owns a Redmi K70 Pro phone?

[sleek vortex]

not me but why lol

[livid mantle]

cheapest phone made in china that has the flagship processor by qualcom the snapdragon 8 gen 3

[sleek vortex]

oh

[feral marlin]

Hi, sorry if the question is posed before ... does perplexity pro writing mode give exactly same response as the base model api?

[tame current]

is it tos to post a porgramming job here

[tame current]

Idk

It doesn’t seem like it

It’s kinda stupid like it forgets context sometimes in writing mode

I’m not sure if there is an option to clear context and avoid follow up questions?

[warm cave]

Wonder if this is a partnership or if they are just using the pplx api

[sleek vortex]

partnership i think?

they had polymarket stuff inside perplexity too iirc

also long time no see, lol

[warm cave]

Oh really, did not know

Saw a YouTube video that said it was the most accurate type of prediction, was curious the types of things people where trying to predict

[sleek vortex]

how does it actually work?

do they like distribute back the money from everyone to the people who said the right thing

but then how do the ones with multiple answers work

[serene sand]

p2p prediction

[toxic charm]

hi

[sleek vortex]

i might be mixing things up

[toxic charm]

are you ml engineer?

[sleek vortex]

possibly got confused in my head about the ceo tweeting a lot with poly market screenshots …

[serene sand]

no

[toxic charm]

anyway, are you developer?

[sleek vortex]

however you can get polymarket graphs inside perplexity

through tako

[serene sand]

I am community mod

[toxic charm]

ah, sorry - I am python dev... is there any thing i can help you??

[sleek topaz]

How much do you reckon it costs perplexity per month for each user model wise? Obviously that’s based on the amount of prompts they make, the model they use and how many tokens each take up (including responses), but given the average user of perplexity, how much do you guys think it would it cost in computing costs for the LLM usage per user roughly

Let’s say each user was using Claude sonnet 3.5 btw just for simplicity and since it’s the best model with the most queries rn

[toxic charm]

sounds good! it is more important to calculate the token pricing when using llm generally..

to be honest, i am not perplexity dev but openai dev.... who can tell us pricing of perplexity simply?

[sleek topaz]

I did some calculations and I determine it would cost them $1-10 USD per month for both average users and super users (users that use tons of queries with massive tokens)

Some super users might cost them a lot more than even $20 a month but they are outweighed by the majority of people who use the service within the profit margin

That’s just for LLM cost though and doesn’t factor in all the other stuff a user would be costing them

but yeah I’d imagine for at least 90% of users it would cost them only a few dollars at most a month for LLM costs alone

What are your thoughts?

[toxic charm]

thanks for your effort, in openai, calculate the token price as a unit of 1M tokens - for example, for gpt-4o, $2.50 / 1M input tokens, $10.00 / 1M output tokens

[sleek topaz]

Oh sorry I was using Claude sonnet 3.5 api pricing

Gpt4o costs less though so that would make it cost less for perplexity for those who use Gpt4o as their main model

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@sleek topaz

[sleek topaz]

Claude 3.5 Sonnet
Our most intelligent model to date
200K context window
50% discount with the Batches API*
$3 / MTok
Input
$3.75 / MTok
Prompt caching write
$0.30 / MTok
Prompt caching read
$15 / MTok
Output

[toxic charm]

i think it's more expensive than openai

[sleek topaz]

Yes

[toxic charm]

what is the good things of perplexity than openai?

[toxic charm]

as i know, perplexity is based on search engine, right?

that means web-based search

[sleek topaz]

Internet access/their sources system, how you can use multiple models of your choice, the new reasoning system (for any model it seems), profile system/collections, cleaner interface and you can see how your prompt is being interpreted by the model. I haven’t used chatgpt in forever but it lacked a lot of stuff perplexity has last time I used it. You can probably configure OPENAI to do some of this stuff through API but the average user wouldn’t know how to do that anyway

And image generation with tons of the modern models

Perplexity has been my go to AI service for around 6 months now and it’s great

[toxic charm]

interesting! if so, i will use it at once... i have a question, are you perplexity's dev?

[sleek topaz]

And how you can focus sources specifically to things like Web, Videos, Academic papers/research etc

[sleek topaz]

You definitely should try it, it’s well worth it. And no I do not work for perplexity, I’m just a low level software dev (not AI related) and perplexity has done wonders for me so far

[toxic charm]

wonderful, i have a thing showing you... i will send it into DM

[supple pecan]

Possibly! Give it a shot, see if you like the results. You can also go to the GPT battleground and try using the same prompt for every GPT and see which answer gives you the best results for you, personally. That's probably the best way to find out which one specifically is the going to be your winner.

[supple pecan]

I for sure like to be able to have my GPTs have "personalities" so it doesn't feel like I'm talking to a robot, and sometimes I like making personas that are just funny. Its not a make or break, but I do like the option

[wide spoke]

https://x.com/Ultra_IA/status/1847821253476008227 NO REAL WTF

UltraIA (@Ultra_IA) on X

LOL

Twitter · Oct 20, 2024, 2:04 AM

[livid mantle]

nvidia/llama-3.1-nemotron-70b-instruct api is so good

[tame current]

Lol wtf imagine if it was real though

[wide spoke]

is real

[tame current]

well I dont believe it to be genuine

[wide spoke]

i just spent an hour talking to a guy on discord about working together and i found out it was a bot.

https://x.com/i/spaces/1mrxmMepOjgxy

Happening now: lily reveals it all 🌸

Lily Ashwood’s Space · 0 listening · Where live audio conversations happen

Twitter

This demon for me is AI and speaks like a perfect human.

[valid iris]

Why can't the user just make their own focus mode the same way DuckDuckGo and brave can use goggles to only search for results from specific websites? @signal hamlet

[tawdry socket]

Looks like Perplexity closed another round at 250M on Oct, 15th

[ashen folio]

Hi

Hi

[versed plaza]

So after analyzing my pantry the AI suggested this meal for me...should I try it?

Chicken Salad Sandwich with Chamomile and Lavender Tea
Ingredients:
Cheesecake Factory Bread: Two slices
Tillamook Seriously Strawberry Cream Cheese
Chicken Salad (one cup): Already has grapes, which will complement the strawberries in the cream cheese.
Chips and Queso for sides or dips.
Tea Pairing:
Chamomile and Lavender tea with honey and lemon

I never thought about putting strawberry cream cheese on a chicken salad sandwich before or seen anyone do it that way... hmm

[cloud sigil]

anyone in school and can recommend some websites that use good ai to organize ur stuff for u? like some websites i saw can make flashcards for u but can also be storage for notes as well

[livid mantle]

nvidia nenorron 70b sucks

im going back to o5-preview

[worn monolith]

revisely.com seems relatively reasonable. I'm not in school anymore, but I remember using the free version of Claude to make flashcards. It did a really good job with some prompt engineering.

[versed plaza]

Use obsidian!

They may even have a flash card plug in

[cursive jacinth]

coco note sir

[tame current]

I’m a college student and I was wondering, could Perplexity help me with my math and science?

[misty root]

What model does the normal pro search use?

[placid coyote]

cool, but very slow, expensive and possibly unreliable calculator 😅

[halcyon coral]

Hey @misty root!
Pro search always uses the selected model from your settings.

[whole hill]

hey peeps, what do people use pro search for? finding normal search has been great for me so far!

[valid iris]

It conducts more than one keyword search, so it is more thorough, but recently, I have been getting less than 10 results for each response, so i don't know if it helps still

[golden anvil]

Really?

[tired fractal]

.

[vast matrix]

Two girls in a. cut

I don't know who to choose

I fell in love with her

I fell in love with you

Two girls in a cut

which one to choose?

[sleek vortex]

hardstone psycho 🔥🔥🔥

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@livid mantle

[misty root]

I'm on the free version, so I can't select a model.

[sleek vortex]

then it likely uses whatever model they have free for you

maybe gpt4o but probably a sonar in house model

[misty root]

OK Thanks! I think it's GPT 4o Mini, because the responses are similar to my personal web search with OpenAIs API.

[sleek vortex]

I feel like they wouldn’t shell out extra capacity for 4o mini

Perplexity operates at a scale where they have to pay for whole units of models over just per token

so it financially makes more sense to just send your query to whatever model is free right now- from the pro selection

so likely GPT4o then, possibly fine tuned

[misty root]

Oh I see. Thanks!

[sleek vortex]

also may depend on your region

e.g pro search off might use different models here compared to say in japan

as they may have seperate allocated compute there for lower latency

[stray seal]

Not sure if this has been discussed before, but it's not possible to use Perplexity app on Vision Pro, and the web version doesn't have the high quality voice input the app has. Using iOS dictation is pretty bad experience.

A dedicated AVP app probably is out of reach but can the team at least port the iOS app to AVP to greatly improve user experience there? AFAIK porting iOS app to AVP should be fairly straightforward (just toggle the option)?

Also, looking forward to see the Grok API integration soon 😍

[odd mauve]

Hi, is it possible to programmatically access not Perplexity models, but the output that is generated based on search when I ask a question on the website?

[plucky pebble]

Do perplexity pro have o1 mini or preview?

[feral jay]

How can I make perplexity use more sources, like it used to long ago, and not cut off at 3-8? Asking for more in the prompt seems to have no effect.

[plucky pebble]

How use it I do t see option

[silent cairn]

How do I get to speak to somebody at plex on enterprise subscriptions?

[wraith crow]

You can send an email to [e-mail]

[wraith crow]

You can use o1 mini by using the reasoning focus

[livid mantle]

hows ur perplexity website project

[pliant widget]

sorry don't kill me but can we share referral codes or get them on this discord or is that not allowed?

[tame current]

is there any way to use multiple images in the same thread? i want to use one image per question, not have the first image used in the response with the second question
@ok.alex

[pliant widget]

if it is then : >

[ashen reef]

Hi all,
Can anyone advise me on what to do when we are not getting the expected context for a query, even though we have the context in the vector store?

[sleek vortex]

slow but

making progress

[agile blaze]

since my upgrade to an Enterprise Pro subscription, I seem to have lost the Focus option. is this normal? All I have are options to choose sources and attach files. I miss my Focus options. 😦

[livid mantle]

https://tenor.com/view/trout-trout-gang-thumbs-up-funny-animal-awesome-gif-25706215

[halcyon coral]

Hey @brazen sphinx!
The students will get the free Pro but it's not automatic. They need to click the offer banner on the left sidebar of the desktop version 🙂

[knotty parcel]

wanted to ask, how do you guys get perplexity to provide answers without looking things up online? sometimes I want to use perplexity like chatgpt or claude and just have it answer questions (coding stuff, mostly) without searching things up.

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@knotty parcel

[minor shard]

Use writing focus 🙂 and turn off Pro search

[rapid wigeon]

What is happening with API? I am getting internal server error and usage asked me to add my credit card again, usage got reset etc. weird

[fervent spruce]

someone put me up to date to why the perplexity.ai focus is non existent?

[tardy junco]

okay whos ready to talk spaces with me

[full gyro]

A Pumpkin has appeared!

Time is up!

[tardy junco]

are there any websites, or good resources where peple share there spaces for collaberation or ideas

similar to discord but like not discord

Someone give me the run down...i dont see anything on the same level as perplexity right now
anyways...spaces. what are some good resources for space ideas, prompt ideas, etc
also is claude 3.5 sonnet still the best?
with perplexity^

[livid mantle]

ChatGPT, HuggingChat, Microsoft Copilot, Google Gemini, GPT4ALL, PromptPal

[tame current]

aw hell nah 💀

im on university wifi lel you already know they be intrusive

[tame current]

That’s not only Perplexity a lot of services do that because of bot

did they do this before?

@fierce sonnet I've confirmed my personal info cuz im a campus strategist, can I have my vpn IP exempt from this?

[fierce sonnet]

hey, reach out to [e-mail] 🙏 we'll take a look!

[tame current]

alr thx

but i cant have AI draft the email for me 😭

[orchid magnet]

been thinking about cancelling my chat gpt membership for perplexity to help me with school, and truly grasping the material.

What has changed in perplexity over the last year

[pseudo bison]

Yow guys

[vast matrix]

don toliver

@everyoneily

@everyoneily

whyd that send twice

Lol

So funny

Anyway

How was yalls day

Oh good

[pseudo bison]

Do you think I can automate bot using perplexity

[vast matrix]

Thanks for asking

[fervent spruce]

Hi, do you guys have good ai prompts for collections for coding and programming?

[cloud sigil]

how good is Claude 3.5 sonnet (free) to like ChatGPT free 4o? What about other models that can keep up?

[kindred folio]

when will remove search suggestions / discovery toggle be added to ios app 🙏😭

[cloud sigil]

I once asked about this thing called notebooklm and it uses Gemini flash or pro? Which one is better and how does it compare to gpt 4o?

Or Claude 3.5

[steel geyser]

I would love to hear from the developers of perplexity about the lawsuit that was filed for alleged copyright right violations. Because, I’d like to know what we should be expecting in terms of subscriptions meaning will we see an increase in the subscription price. Also, does our subscription payment go towards the legal fees? These questions are legitimate questions and call for transparency.

[burnt cliff]

🎯 Join Hack for Social Impact November 9-10 in SF and tackle real-world challenges through innovative data and software solutions! We've teamed up with nonprofit and government organizations both Bay Area local as well as international, including the United Nations Convention to Combat Desertification to bring you pressing societal challenges that are well-scoped and technically creative. These problems can be solved by you! Here are some examples:

​💧💧Land degradation, drought, and people
🇺🇳🌏💬Fast and accurate translation of UN documents in highly technical contexts
​🏠🏠Understanding building codes for housing affordability

Besides cash prizes and perks from our sponsors, winners will get the chance to implement their ideas with our partners and drive meaningful change. Register here: hackforsocialimpact.com

[rustic orbit]

why reasoning mode does not search websites?? I input my queries and even use "search the paper" but it does not search the web, why it seems so different to the announcements demo?

[shut light]

do you have perplexity pro?

[rustic orbit]

yes i do

[latent acorn]

any o1 updates?

[shut light]

hmm weird, i got no idea unless its something slowly rolling out

I did the same question as in the demo video and it seems to be there

[rustic orbit]

is it not support multiple questions at once? I input 4-5 questions

[shut light]

I don't know, maybe try one and see if it works @rustic orbit

I don't see why multiple wouldn't be supported though

[harsh verge]

#1297990620216037450 message

[supple pecan]

lol they aren't going to issue any statements. Instead they silently upgraded their "terms of service" or whatever, and basically just said "we'll do better we swear". Then I was able to get perplexity to tell me that it was not following those rules after I copied and pasted them to it. It even acknowledged the existence of the update, but I have an entire thread of perplexity breaking the 4th wall and spilling ALL the tea. Crazy stuff haha

[fervent sundial]

Does perplexity generate ppts?

[fast meteor]

no

not directly

[mint forge]

Hi all

[cursive jacinth]

threads from perplexity broken sir

[rustic orbit]

I only input 1 question but it does not work too, and it likes do no have the thinking process, it just output after I input the query, no any latency

[fervent sundial]

I understand

[obsidian crystal]

yuck, gross… I hate when websites do this

They also block certain websites at my job.

[hexed rampart]

Hi all! Is there anyone here who knows about the 'closed beta access'? I applied months ago and still haven't heard anything. Most of all I would love to try the 'return_citations' flag in the API..

[elfin portal]

me looking around for my org that I now have

(its just me)

[vapid onyx]

Still cool that you can add internal sources

[calm compass]

Since grok has api are they adding it to perplexity??

[valid iris]

Perplexity AI: What percentage of organic chicken processed in America is washed with chlorine? https://www.perplexity.ai/search/what-percentage-of-organic-chi-xeWXj9qUS7i.AA3Kl3Vk.Q

[full gyro]

A Pumpkin has appeared!

Time is up!

[valid iris]

I specifically asked it to search for results from usafacts.org, and not a single result came from that website.

Could that be because there were none?

[versed plaza]

Do collection system prompts have a 1500 character limit on mobile but no limit on desktop? This is how it appears.

[dim wind]

https://x.com/AnthropicAI/status/1848742740420341988

Anthropic (@AnthropicAI) on X

Introducing an upgraded Claude 3.5 Sonnet, and a new model, Claude 3.5 Haiku. We’re also introducing a new capability in beta: computer use.

Developers can now direct Claude to use computers the way people do—by looking at a screen, moving a cursor, clicking, and typing text.

Twitter · Oct 22, 2024, 3:06 PM

AYO

[livid aurora]

Guys https://x.com/AnthropicAI/status/1848742740420341988

Anthropic (@AnthropicAI) on X

Introducing an upgraded Claude 3.5 Sonnet, and a new model, Claude 3.5 Haiku. We’re also introducing a new capability in beta: computer use.

Developers can now direct Claude to use computers the way people do—by looking at a screen, moving a cursor, clicking, and typing text.

Twitter · Oct 22, 2024, 3:06 PM

[tired fractal]

is this already live on perplexity?

the upgraded 3.5?

[dim wind]

it just come out I think so idk

[halcyon glacier]

Hello, would it be possible to create larger "Spaces" to explain more stuff?

[dim wind]

it would be cool to get an update on that

[halcyon glacier]

Yes

[slender imp]

I would like to know too!

[minor shard]

Yaayyyyyy Claude

[prime sigil]

With perplexity API , how to get in return the URLs of sources and further reading, I have tired this in the prompt " Very important - Include sources and further reading full URLs address in a separate line per each URL." but is doesn't bring the URLs, sometime it brings name only.
Have anyone tried this? Thanks

[tame socket]

new claude

When will sonnet be updated?

[placid coyote]

what copyrighted material? am I too uncultured to know what sonnet means?

[cinder comet]

soon

[opaque stump]

LFG.

[tired fractal]

LFG?

[cinder comet]

is it?

how did you know?

is it time for me to resub?

[opaque stump]

I’m saying let’s get on it

Not that it’s here

[manic kestrel]

beep bop boop, my ai is being stupid $&@!
I ask pplx-ai pro and got this (truncated here) The phrase "Scrolling like it's 1999" appears to be a modern adaptation of the more famous expression "Party like it's 1999," which originated from Prince's 1982 hit song "1999". However, the specific quote "Scrolling like it's 1999" is not directly related to Prince's song or any other well-known source...

[chilly fox]

Looks like they are adding both new models when they are both out

[tired fractal]

ye but when>?

[torn sierra]

I guess it should be on today - considering previous upgrades. With the exception of o1 ofc 😉

[frail parcel]

How does it decide internal knowledge sources? I’m just using my personal account, but I’m at work, but I’m not aware of any perplexity usage here

[torn sierra]

It's an add. Just close it 😉

[fathom kindle]

Can I get this fixed?

Why is this such a problem? How do I report this to support? Why is it so god damn difficult to contact support anyway?

[placid coyote]

toying with ui thingy, I was quite happy sonnet managed to use an image from (maybe) search results. very hacky and barely working. but sometimes, rarely, it does something 😁 (no, it doesnt generate images on its own and because of the way of simplest implementation I could think of, sonnet has a lot of trouble utilizing it)

[pure bay]

Maybe create a post in #1140622008086970420 without the swearing.

[chilly fox]

Guys, Perplexity has some competition, Yahoo has added a AI chat service (I wish I was joking)

[cinder comet]

do we have the new model yet?

[raven yacht]

Unless they have API credentials im not to worried

[tame current]

Just want to say I freaking love PAI! I use it all the time for work and personal. And it always comes through with great information and the sources right there!

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@crystal jay

[torn sierra]

This pro engine could generate images using the selected model in Perplexity. If it can handle Python execution, why not an image generation using e.g. Flux?

[livid mantle]

This pro engine could generate images using the selected model in Perplexity. If it can handle Python execution, why not an image generation using e.g. Flux?

[kindred folio]

i’ve always thought that change app icon being in this menu is the most useless thing ever. people usually change their app icon once and there’s an option in the app for that

[chilly fox]

New Sonnet is here!

[old blade]

Soo weird. i was doing research and i releasied he switched his behavior. after i reallised it i checked Twitter and they really changed the model. Its kinda crazy because he starts asking questions and gives multiple options how it should do research. im feeling weird right now hahah

[kindred folio]

is it on mobile bc it doesn’t have the as of oct 22 description, but there was an app store update for perplexity so idk

[chilly fox]

It is not on mobile yet from what I have noticed but that can change very fast the speed they work

[tardy coyote]

Hello ! If I choose a model like GPT4o if I do quick search and I ask “are you chatGPT” sometimes he says “yes” sometimes he says “no I am not”
Are we sure he always chooses the model for every requests ?

[keen vortex]

Woot, Sonnet 3.5 makes my self-recursive framework self-aware 😄

[plucky plume]

Is perplexity getting new sonnet 3.5

[plucky plume]

Damnnn

[subtle ridge]

So how the response made by new sonnet?

[opaque shoal]

I'm curious—what's keeping the new reasoning mode from being brought to the mobile apps?

[sudden timber]

hi

[tame current]

is pplx using the new sonnet?

[tame current]

Yes

[tame current]

i want to use perplexity pro for multiple choice data structure questions.

would you guys recommend having pro on/off
what type of focus? writing or web?
and would you guys recommend changing anything in settings to make it more correct?
or prompts?

[keen vortex]

Uhg, new Sonnet sucks, there needs to be an option to switch back 😦

I use Sonnet for writing, and where I used to get 600-800 words with my writing prompts, I get around 80-100.

So that 10x's my workload and prompts used, at least. Since I have to create separate prompts and stitch things together post hoc.

And fiddle with prompts to get things to flow properly, redo sections that don't flow properly, etc.

It's reasoning is a little better, but not enough to make up for the throttled max output length.

And has anyone noticed it doing the 'I am not comfortable about talking about that' about every damn thing?

[autumn lark]

Yeah old Sonnet please. Revert.

[keen vortex]

I'm waiting for the new sonnet to say "I'm sorry Dave, I'm afraid I can't do that." 🤣

[rapid nymph]

seems like system prompt isn't working, anyone noticed?

[keen vortex]

If there's no option to switch back to the old Sonnet 3.5 in Perplexity I am going to have to switch to APIs to use the old Claude Sonnet 3.5 with a services like Jina.AI and Tavily for web search. RIP Perplexity sub. 😦

[last steeple]

it seems that response of Claude 3.5 Sonnet in Perplexity doesn't same as claude's api in Poe, the latest knowledge until October 2023

[keen vortex]

Perhaps the devs could just add two sonnet 3.5 options in the dropdown, one for the old one and then the new one with 10/22 date tag on it?

[keen vortex]

Hmmm, and the new Sonnet 3.5 on Claude.AI says it's cutoff is April 2024.... I wonder if the models are hallacinating when asked this?

[rapid schooner]

Directly asking models about their training cutoff dates is often unreliable.

[cedar grove]

Why can’t show the specific formula form

[rapid nymph]

you mean the LaTex?

[alpine pike]

just paste that into a latex compiler

[cedar grove]

oh，I solved that,thanks

[halcyon coral]

Hey @last steeple!
Perplexity uses a different system prompt, please check https://discord.com/channels/1047197230748151888/1225549032718602332

[tame current]

cool

is there a way to type to perpolexity on apple watch

[summer halo]

I think there is something wrong with the AI Prompt share in spaces....it does not seem to acknowledge my preset prompt...is anyone else facing this problem?

[dusky linden]

If you speak or query it in english. Iam bi lingual in german and english and the result all depends on the input language or the keyword "translate"

[rapid nymph]

Yeah, and I have this problem not only in Spaces, but also in any threads. However, it works once I enable Pro mode.

[dusky linden]

I recently asked to find me lyrics of a phrase in a song I was unable to identify and it just conjured up some random artist name and soong out of its imagination. So results from perplexity AI in general should always be back checked

[frosty stirrup]

does anybody wanna join my online class

[carmine hare]

I too have encountered this problem, now I have to write a long prompt every time

[high lake]

Is there a timeline for when perplexity will get the (horribly named) claude sonnet 3.5 New version?

Ah nevermind, scrolled up in the chat :'D

that was quick!

Another question, do we get to see what the system prompt is it uses for the different focuses? Especially for writing mode where you're just interacting with the LLM without web search?

[rapid schooner]

This problem has persisted all day. It seems that custom system prompts within Space aren't working.

[balmy flint]

yea, that why my prompt is always end with "please based on information in instruction" , I even set the key shortcut for it 💀

[tame socket]

What specific time will the desktop app be released on macOS?

[versed helm]

I accidently deleted a thread today on iOS by swiping, can it be restored by support if I provide details?

[stable notch]

Claude was giving me pretty good answers and now he's giving me horrible answers, but I don't jump to conclusions, I use him for roleplaying, has anyone else noticed that his answers are too stupid now?

[torn sierra]

The custom system prompt is not working either. It is working only with pro on. When pro is off it’s not working now 😢 I hope it will be fixed soon.

[tranquil orchid]

Hey @torn sierra!

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.

[left stream]

Anyone know how I can go back and see the beginning of a thread? It was for writing but the original beginning of the thread isn’t visible and half of the thread is showing

[tired fractal]

It's a known bug I reported ages ago, and the devs don't care. Only way you can view it is by deleting oldest messages, and for every old message you delete, another message at the beginning of the thread pops up.

If you have been writing for a long ass while, it'll take a lot of deletes before you see beginning of the convo.

So, really it is an UI visual bug, as the text remains. A fix that wouldn't take much time.

But again, devs don't care, so...

[left stream]

Damn it

It was a solid 4 hours

[keen skiff]

uhh

where do i complain that I had an year of subscription but somehow i dont have it anymore?

[keen vortex]

Why was the old Sonnet 3.5 model removed rather than just adding a new model entry ie: Sonnet 3.5 10/22 and stick a [New!] tag on it and rename the old model Sonnet 3.5 Originals or something like that instead of removing a best in class model for numerous use cases? The new model is slight upgrade in logic and reasoning (useless for my writing) and a massive downgrade for content creation. I use Sonnet 3.5 for academic writing, and where I used to get 600-800 words and dozens of verbatim scholarly quotes with my writing prompts, now with this new version I get around 80-100 and maybe 2-3 quotes if I am lucky. I used perplexity for my writing since it could search academic databases for the latest research and with the old Sonnet 3.5 could output writing ready for my use case with minimal editing and which I just had to verify the academic sources to ensure the entire writing was accurate. So that's easily 10x's my workload and prompts used, at least. Since I have to create separate prompts and stitch things together post hoc and fiddle with prompts to get things to flow properly, redo sections that don't flow properly, etc. I've paid for a year subscription, but now the service is basically useless to me... I will have to switch to using Sonnet 3.5 through OpenRouter's API and use other APIs for academic database searches, which means hoping I get a refund or I'll have to initiate a credit card reversal as the service is unusable for the use case that I purchased it for, and which it previously worked for. @toxic moss

[tired fractal]

Given the way AI changes from one weke to another, a long-term sub like a year is a bad call. One year is a long time for AI development, far more useful stuff can be made, or what's currently useful may no longer be so. Stick to 1 month subs.

[keen vortex]

While that may be true and I don't disagree in principle, this was entirely avoidable and unnecessary. I pay for year subs because they are cheaper and if a service makes a shortsighted change that makes it unusable I have had no issues with my CC company filing chargebacks and refunding me.

[minor shard]

I believe they are just following what Claude's latest model is. Unless you use Claude's API, I don't even think you can use the older version there either.

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@minor shard

[tame socket]

What specific time will the desktop app be released on macOS?

[keen vortex]

That wouldn't apply to a provider like Perplexity, they either self-host the models or API call other providers, original Sonnet 3.5 is available from countless API providers, including Anthropic.

[keen vortex]

Perplexity doesn't use Claude.AI website, that's a separate product from Claude's API.

And if Perplexity is self-hosting Claude 3.5 then they can run both side by side. Either way it would be just as easy to add a model to the dropdown as it would be to replace it entirely.

[minor shard]

I understand that, but their way of doing things has always been to update the existing product to the latest. They did it when they switched from Claude 3 to 3.5 and GPT4 to GPT4o. The old one doesn't stick around. This is not something new with PPLX.

[keen vortex]

Continuous oversights don't make them less oversights, especially when the removed model is still best in class for many use cases.

This is the exact oversight I am trying to address.

[minor shard]

I mean they could include every API from both Anthropic and ChatGPT but that is a lot to maintain.

[keen vortex]

Let's debate about how we can make the service better, rather than continuously making oversights? Also I didn't argue for that, you're misrepresenting my argument to make it easier to attack. Keeping one model revision back when it's still best in class for most use cases isn't what you are suggesting, but it is what I am suggesting.

Adding a model and a new entry to the dropdown doesn't create any more maintenance either, that's a non-sequitur fallacy, and not how anything works.

Once it's setup, it's setup. The only thing they did in addition to setting up the new model is remove the dropdown entry and remove the api calls (whether local or external).

[minor shard]

I wouldn't agree with this as we do not know how they are handling the API.

[keen vortex]

Yet your argument is based on your assumptions about how they are handling the API? That's a no go for me.

[minor shard]

No my argument is if you did not know that they stick to the latest model thats on you. Not them. Its a good suggestion though to keep a previous version.

[keen vortex]

Anyways, I will wait to hear from a Perplexity team member, as this is unproductive to keep going back and forth.

[keen vortex]

You've convinced me I should file my CC chargeback, if that was your intent well done. 😛 In reality though, it's not on the consumer to intuitively know a company makes shortsighted decisions until they experience them. I'd have no way of knowing that they instantly remove best in class models for no apparent reason when a new limited side-grade (or for many use cases a downgraded) variation is released (notice how Sonnet is still 3.5?). Your arguments are like goalposts that move out of the way of the ball, lol.

[keen vortex]

This was a fun test of my critical thinking AI framework. You might find this helpful. https://chatgpt.com/share/67192851-da20-8009-a5dc-c819161d81e9

[minor shard]

Hey I just assume you do your research before investing in a product. Responsibility is as much on the consumer as the company.

[alpine pike]

this server has made me lose faith in humanity

[placid coyote]

hmmmm, why one, and only one, of my spaces no longer injects collection/space pre-prompt 😑

[keen vortex]

Yes, let me go research what a company will do in the future, which they don't have mentioned anywhere on their website, in any of their public documentation, or anywhere else on the internet. Sound plans there boss. 🤣 Keep repeating the blame shifting fallacy, I am sure it will be productive for you. At the end of the day it's the company's job to communicate their future plans and changes to their consumers, so their consumers can even have the 'research' available to find, this wasn't the case here. Are you suggesting I should have spent hours searching social media for user comments about this, and assumed they were correct? You're making zero sense, and arguing against making a product you apparently use better. You can do better than this.

[placid coyote]

oh, the pre-prompt in spaces doesn't work when you don't enable "web search"? 🤔 eh, why?

[keen vortex]

Yes, I ran into this as well. 😦

Sometimes it works in a new chat, sometimes it does not.

[harsh stag]

I thought I was going insane, but apparently it's a regression

[icy cloak]

what are people using for their default llm? gpt4o or sonnet3.5 new?

[harsh verge]

https://x.com/AravSrinivas/status/1849134236525347225

Aravind Srinivas (@AravSrinivas) on X

⌘ + ⇧ + P release version tomorrow. Will start off as your shortcut to asking anything without opening a Chrome tab. And we’ll be investing more into local desktop oriented productivity improvements!

Twitter · Oct 23, 2024, 5:02 PM

[keen vortex]

GPT4o has web search now on the moible app? 😄

How do I get this? 😮

[icy cloak]

this doesn't answer the question at all rofl

[keen vortex]

I'm using GPT4o because the new Sonnet 3.5 thinks that 1800 word prompts requiring deep academic data analysis and full narrative explanations should be responded with 80-100 words and without most of the elements requested - though GPT4o isn't great either.

But hey, the new sonnet has memorized how many letter R's there are in the word strawberry and a few arbitrary logical tests with humancentric framings and anthropomorphic assumptions, so it must be better right?!

[stable notch]

I have two Pro+ accounts, but in one of them Sonnet responds to me as if it were a GPT chat, with silly answers, in the other it responds differently, is there any solution?

It doesn't just happen with Sonnet 3.5, even Opus responds like this, it gives me horrible responses, it's as if all the models were GPT chat and it's frustrating

Help me please

[tame current]

I'm so happy with Perplexity reading XLSX!

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@serene sand
@sleek vortex

[serene sand]

https://x.com/testingcatalog/status/1849191714134843612?s=46

TestingCatalog News 🗞 (@testingcatalog) on X

"Buy with Pro"

This is how y'all are gonna shop in 2025 👀👀👀

Twitter · Oct 23, 2024, 8:50 PM

[alpine pike]

honestly for that stuff better to write your own rag engine locally

I have one I use for all my uni courses to query all the textbooks and past exams and whatnot

uses llama 7b running on my mac and llama index for the rag part and it's surprisingly good

[sleek vortex]

currently designing a complex rag system

like perplexity's but actually good (is the plan)

[alpine pike]

ohh is that the website in your bio

it's super impressive

have to say

[sleek vortex]

the website is quite behind interms of updates

been doing lots of stuff locally

probably will ship it once its all more refined

[sleek vortex]

e.g i did follow up messages but need to do this for it to be good ^

[livid mantle]

I know

[tired fractal]

ngl I'm expecting u to say m'lady and tips fedora every moment now

[alpine pike]

awesome!!

looking forward to seeing it

[keen vortex]

Yes, I do have OpenWebUI and n8n setup, and it looks like I will be using them more and more as Perplexity seems to not have any interest in usability or unplanned 'upgrades' that break stuff (web search in spaces most recently when it was working fine with the previous setup), and removing best in class models for most use cases for cheaper to run 'newer' versions.

And they seem to frown on trying to talk to @signal hamlet, @fierce sonnet, @toxic moss. Apparently the team members are for looking at rather than communicating with.

[raw depot]

when will we have the new Claude 3.5 sonnet ?

[tame current]

Woah. I didn't know Perplexity had their own server on Discord which I'm new to. But I'm glad to explore. Hey everyone

[visual anvil]

Where on the perplexity app? I already got it

[raw depot]

damn

I didn't check the website

but not on the app yet

how's it?

[visual anvil]

I guess so - the app need an update

[raw depot]

wait let me update

updated

Idk, doesn't say the new 3.5

[visual anvil]

If they do like on web.. they’d have to re update the app to take off the new tag.

[signal hamlet]

Hey @keen vortex! Thanks for your feedback, but sooner or later we've always updated the models to the latest versions, that is the updated Sonnet in this case - the most advanced one, according to Anthropic.
If you have examples of something that is not working as it used to, please file a bug report with thread URLs, so we can investigate and reproduce the issues.
If you are referring to the writng mode switching to search - that bug has been reported, and our team is aware of it and working on fixing it. Thank you!

[rotund geyser]

It's still using the newest model even if the app doesn't say so

[keen vortex]

So no chance of having the other version of the model that actually works properly as an additional option? If so where do I request my pro-rated refund, or should I just not bother and file a CC chargeback?

I am fine with either option.

[keen vortex]

And no, I am referring to the preprompt in not working in the 'Spaces' when web search is enabled.

[signal hamlet]

We use the latest models as they become available, and I'm afraid we'll go on with the latest versions.
We've received reports that AI prompts aren't working properly with Pro off, and we're investigating this. If you have specific examples of issues, please let us know and we'll take a look!

[summer crescent]

I had several instances yesterday where both pro and non pro would tell me it was incapable of searching ....

[summer crescent]

it also hatched on to an image I had uploaded and absolutely refused to focus on anything else but yeah

[tame current]

I just sold my soul to a ancient hindu god. I told Lord Hanuman that I'll serve him 1000 years in the afterlife if I wake up 6'9 tomorrow

[versed plaza]

How much of perplexity does Jeff bezos have a stake in?

I know he gave anthropic like 500 million dollars or something

[minor shard]

u should ask pplx

[versed plaza]

That is definitely normal perplexity flaw with any llm. I had a little success by linking images instead of uploading. But sometimes it says it can't see them at all so I'm not sure if certain models can view jpeg links and others can't?

[versed plaza]

Can't you just use opus? I thought that was better than old Sonnet anyway

[left stream]

Do you know how to see the beginning of a thread? I was writing a lot but it cut off in my space and I can’t see the beginning of the thread anymore whether it’s on the website or the app, it only shows half of the thread

[summer crescent]

Gemini before 02 would do the same with any file you gave it, latching on to it like it's life depended on it and regardless of how you tried to get it dismiss the file it would keep focusing on it.

While the ui shows it keeps re reading the documents regardless with the new model it's atleast capable of focusing on your actual question and capable of ... After spending resources on re reading that the document wasn't relevant to that exact question.

So I suspect it's a llm thing in general in regards to files.

Master gave dobby a picture!

[versed plaza]

Whatever it is it definitely makes it hard to use one conversation for a project.

As it is I only upload a picture when I'm prepared to close a conversation out afterwards cuz past that point it's better to just go off of no context at all

[summer crescent]

Yeah I get around the problem in gemini by uploading the text in a pastebin and then providing gemini the link.

Your linking pictures seems to be much in the same nature

[versed plaza]

Okay sounds like we both figured out similar work around. It does seem to perform better that way but sometimes I wish I could throw out old context from a conversation completely

Especially if the conversations main purpose is to walk you through various steps of something

[summer crescent]

Well to be fair, it's context isn't all that long.
But I get what you mean

[versed plaza]

It's hard for me to understand how it trunciates context once it hits the window

[civic fog]

W chat

[tame socket]

Make chatgpt great again

iOS Siri also needs a ChatGPT subscription.

[solid star]

I was testing out different prompts and found one that makes the responses insanely long. I got the idea because I remembered gemini would be limited in response size by the google app itself that had a prompt to make responses small since the longer it is the less interested users would be.

Works insanely well

[minor shard]

you can use it without the subscription. Just won't be as good, uses the free plan without sub and also you can use it without an account and it also uses free plan.

[karmic spear]

did any of yall get the app

mac app

just got it rn

[tame socket]

mac desktop Released

[gritty turtle]

is there a windows app aloso ?

[tight quartz]

yea but the shortcuts conflict with other apps so I have the app closed for now

[severe crystal]

I just downloaded the app on MacOS. Looks all nice except O find the font size too small and there is no way I can increase the font size of the text in the side bars and the main window when the answers appear are too small. The app does better in a browser. Please fix this and at least give us the ability to increase the font size.

[cunning tendon]

i am unable to login to the mac desktop app with the credentials that work on the web. appreciate any advice...

[versed plaza]

Perplexities space editor really really loves closing itself out when you were mid-edit.

This has happened at least 40 times to me today

I think it must either hate the control key or the mouse wheel or something

But even clicking on the emoji button I can't even click through emoji categories without it just closing the interface right out

I was able to reproduce it okay it happens when you're selecting text and you let go of the mouse button while the cursor is outside of the text box even one pixel it just automatically closes even though you didn't click away it treats the letting the mouse button up as though you clicked out of the interface.

So there's at least two things that closes it out prematurely. Emoji categories and letting go of the mouse click when it's anywhere outside text field

Sorry probably should have thrown that in bug reports I have never reported a bug before so not a habit

[keen vortex]

Would you mind sharing? That would be great.

[full gyro]

A Pumpkin has appeared!

Time is up!

[past basin]

damn. perplexity voice has been down all day.

[hardy moat]

Anyone knows if it's possible to preselect a focus when using "https://perplexity.com/search/?q=%search_query%" ??

[cobalt quiver]

Same problem here

I have written support team looking for any advice

[muted pine]

Any possibility to use Pro technologies with API ?

[marble ice]

Is it just me, or does anyone else feel like the new Mac app is poorly written? It uses 20% CPU even when it's not doing anything, and unlike ChatGPT, you can't keep it in the menu bar. As a result, it’s always stuck in the dock.

[vital sequoia]

there is a bug in the mac app

[long flame]

After changing collections to "spaces", all my collections with custom instructions stopped working the way they should. They stopped listening to prompts and I can't yet figure out how to get them to follow them again. Has anyone else run into this? I can't get a controllable result from a customized AI - agent that was previously a "collection"

[old blade]

External mics are not working in Perplexity mac app anyone noticed that?

[split karma]

Perplexity for Mac High CPU usage 🥲

[woven tartan]

Yes i agree 17 or 20 %

More than chrome or disord 😉

[cinder comet]

eh?

is it based on electron?

chatgpt was built using swift

native mac support

[cedar orbit]

@tranquil orchid explique moi la membrane plasmique

pls how to use it ?

@everyone

[cinder comet]

eh

can you write that on google

it gave me a 3d model

is this new?

[cedar orbit]

pls how to use the bot on the server ?

[cinder comet]

there is no bot to use

[cedar orbit]

bruh

thx

[vocal canopy]

I am trying out the MacOS App now:

• How do I hide the Dock icon and only keep the Top-Menu Icon?
• Why do we need all the empty Menu Items when the App is in the foreground?

[manic kestrel]

@tranquil orchid congrats on launching the macos app, for what it does, it seems to work well. However, I don't like that app settings are some kludgy non-standard implementation, which doesn't support the default macOS settings shortcut; also when is image generation support going to be added? TBH, I'm not sure I want to switch from the web version for this version unless there are frequent updates (like once a week) that work towards a feature complete version.

[pearl ferry]

hello, anyone has implemented perplexity in Salesforce?

[versed plaza]

is anybody up bored and feel like giving me feedback on a prompt i'm working on?

[full gyro]

A Pumpkin has appeared!

Time is up!

[tacit meteor]

Mac version font size is crazy... Should I buy glasses to use mac version?

[hard veldt]

Same here. Annoying indeed. It works with internet search only

[tranquil crest]

i do have payed to use the pro but is not code

i want my mony back

[long flame]

maybe you can tell me what's wrong?@signal hamlet

[lapis citrus]

Anyone having this issue?

[half venture]

Are there any plans to add gpt o1 ?

[minor shard]

They have o1-mini. It is the Reasoning focus.

no o1-preview unfortuantely

[half venture]

Is there any limits ?

[minor shard]

Only 10 uses per day I believe

[half venture]

Strange other subscription wrappers can offer o1 preview but not perplexity

[minor shard]

Yeah, o1-mini is good for like Coding. I prefer preview for research

[half venture]

Maybe aravind thinks o1 preview is not worth the price and reasoning

Or maybe just waiting for o1 full

[minor shard]

maybe

[half venture]

What's strange is

O1 preview while being expensive yes

Is still cheaper than the orignal gpt 4

And they were able to offer orignal gpt 4 daily

So I feel like this weekly limit is artificial

[minor shard]

Yeah just to avoid abuse

I wish there was a way for them to allow the non-abusive users to take advantage of a higher limit

like VIP

[half venture]

I guess I will stick to the wrapper then for o1 preview

Since it has no limits apparently

[half venture]

Eh

I don't think it's for abuse

I think it's to prevent competitors

Accumulating datasets

[minor shard]

ah interesting

[open stream]

Are we allowed to put classifieds/jobs in here?

[solid star]

Sure dms

[left stream]

On the api?

Time to fund my account then

[half venture]

Orignal gpt 4 was 20 times more expensive than gpt 4o

[left stream]

Holy

[warm rose]

Do we got new claude?
Antrophic really doing something, the fact it close with o1 is crazy

[warm cave]

Yep we have the new sonnet

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@warm cave

[warm rose]

That good news

[warm cave]

Yep, always a good day when we get access to a new Claude model

Always more hyped than with an OpenAI release

[serene forge]

intresting how they able to throw punch on o1 preview weight class while costing the same as previous sonnet.

has antrophic increase their api price?

[warm cave]

Yeah, it’s great

[warm cave]

I think it’s unchanged, but I would have to double check

[serene forge]

on openrouter sonnet costing about 15$/permillion output where o1 preview cost 60$/permillion output.
actually it suit their change from non-profit to be for profit haha.

[warm cave]

Haha, yep and o1 is way more expensive then a regular model costing $60, it produces soooo many hidden tokens you have to pay for

[delicate glacier]

I did request a refund and saw the refund transaction shown on my credit card, however, I reordered the subscription on the same day, which charged me twice, but the system cancelled my refund.

[grand finch]

how van I use my Uber account to access to Perplexity pro?

[warm cave]

Same riddle

Why did I just spend 10 cents on a stupid test 💀

[dense birch]

why perplexity have become so dumb? i used to translate text to read novel, now it become useless as it start output at beginning, the result shows fine, but at near the end of output, the text become gibberish, with lot of random words throw out. Is there another alternative? there is so much regulation, i just want to quietly read my raw novel, that's just what i want.

and it is too strict with response "I'm sorry, but I can't assist with that. " even though i didn't even give nsfw text at all, just some romance thing, it still can't be allowed...

[serene forge]

intresting, even at the same output token that is 271. it will cost us 0.0165..... using o1 preview, btw i don't include the input on that calculation.
comparing it including the input, basically the total of it resulting in cost of 0.022... for the same token produce with claude.
so yeah, even pertoken OpenAI have more premium pricing.

oh yeah, the model parameter could also be a reason of that.
if i not wrong sonnet have like 100 something billion parameters.
idk about o1 preview tho.

[tame current]

do you guys recommend pro on/off?

[gilded zealot]

I am sure that the Mac app store version is going to be get better over time, but right now it is a bit of a cpu hog while doing nothing, for some reason. The other one in the list is the electron app made for perplexity by webcatalog.

[dawn fractal]

last time my favorite pharse in quote at end is something like
"give me only * [something]* without your commentary"

[serene forge]

what language it is in?

Hermes 3 quite good at chinese and japanese where it also quite uncensord.

[dense birch]

i used it to translate chinese, korean to english

[dawn fractal]

yup i also feel is dumber in recent patches

[dense birch]

is this using offline models? is there an online service... because my pc kind of struggle with offline

also i realize now in perplexity all models are use in GPT4

because all models have giving me response speed same as GPT4 suddenly

[serene forge]

yes, you can get their API throught third party provider.
you know sillytavern? you can use it to be the frontend.

[serene sand]

https://x.com/aravsrinivas/status/1849485216349622462?s=46

Aravind Srinivas (@AravSrinivas) on X

Perplexity MacOS App is now available to all on the Mac App Store!

Twitter · Oct 24, 2024, 4:16 PM

[warm cave]

That crazy, Sam Altman was really full of it when he said intelligence is almost too cheap to meter (of if I remember correctly)

[tame current]

LETS GO PPLX MACOS

[rose spindle]

how to add spaces on iphone app

[tame current]

how is it already number 1 in productivity

[half venture]

Does perplexity have a image filter

[livid mantle]

to save costs

since jeff bezos pulled out

[tame current]

the ui clean asf

[half venture]

He did ?

[livid mantle]

what good is the armor with no brain

[orchid siren]

is anyone elses perplexity this tiny?

[tame current]

grey image

[orchid siren]

u can see it in the top centre

thats the website xD

[nova idol]

whats the advantage with the macOs app over just having perplexity in a safari standalone window?

[tame current]

more convenient cuz keybind

and it stays on top of other windows

[kindred yew]

Anybody know if Perplexity for Mac will get the Enterprise Pro features as well? Have to go to browser to do anything

[buoyant thistle]

can Mac version global shortcuts be disabled or changed? It is conflicting and overriding shortcuts from other developer apps - so annoying that I had to uninstall Perplexity completely

found it - hidden in a "tips" corner and not in settings (or app pereference, where it should go)

[ashen flax]

Anyone find perplexity on mac app has the weirdest font ever. Its hard to read my 50 slides summary with that font

Also it wouldnt let me upload files during the answer. I have to create a new thread

[wicked zodiac]

Has anyone tried transferring your Pro account to a new account?

[balmy dagger]

Ask Pplx 🤷‍♂️

[livid kite]

I'm guessing the hotkey popup

[livid kite]

Ouch! Is it constantly hogging CPU in the background?

[lament imp]

Are you able to run LLMs locally?

Also keep in mind how LLMs work, if you're text is too long, the output will become incoherent because of context length limitations.

Also API ^

[pearl cipher]

yep, cick ont he shortcut and change it

note that they need to have command in them somewhere

lower left cornere

[full gyro]

A Pumpkin has appeared!

Time is up!

[dusky dirge]

Very nice MacOS App. One question: Does the voice chat feature always use the free model even if logged in as a pro user?

[serene sand]

https://x.com/testingcatalog/status/1849506668100395310?s=61

TestingCatalog News 🗞 (@testingcatalog) on X

According to the latest info, we may even see this already in November. Black Friday prep? Will be huge 🔥

https://t.co/B9ZbyNZfSy

Twitter · Oct 24, 2024, 5:42 PM

[wooden bough]

now i want to use perplexity to search this discord for ideas of how people are using it

[gilded zealot]

That’s what i see. A constant 18%

[flint scarab]

is there night mode for the app

[minor shard]

imagine if you could use perplexity to search this discord lol

copy paste is not very convenient

[serene sand]

deep dive ➤ #💬│general message

[viscid juniper]

Hi friends, somebody know if there are some promotion code like Magenta or anything similar, to be able of try the services of PerplexityPro for a friend which she is studying DAM currently? I have my subscription, but she don't have money now and she wanted to try Perplexity Pro for a really good extra tool for the course

Thanks in advance! 🙏🙏

[flint scarab]

is this a pro thing i cant find in settings

[wooden bough]

some of the themes remind of the California Swag 2013 & skating vibes. Tropical and all

[flint scarab]

am i missing something

[wooden bough]

Are there plans to be able to have a.i. suggest you on how to group your threads into spaces? Perhaps group your threads into pre-existing spaces you've made?

[plush fjord]

perplexity desktop "Perplexity quit unexpectedly." when opening it

[serene sand]

I am not using Android, everybody has it

[barren aurora]

COPILOT??? Where are you? Bug report... Copilot is MIA

[flint scarab]

mac app, sorry for confusion

[civic apex]

Windows app when (pls)?

[teal aspen]

I just uninstalled lol

[rustic pasture]

https://techcrunch.com/2024/10/24/they-wish-this-technology-didnt-exist-perplexity-responds-to-news-corps-lawsuit/

[karmic mural]

Who here uses pro? Is it worth the annual? Or how are you using it vs other tools? I'm curious if this is a worthy add. I'm currently using Claude, almost exclusively, and some gpt. I'm now looking to get into the next level and use the tools more actively and advanced

[serene sand]

deep dive ➤ #💬│general message

[karmic mural]

This is a great intro. Thanks @serene sand

[cloud perch]

Does anyone know of a guide that would allow me to learn how to create collection prompts? I'm seeing a variety of styles but having trouble figure out what actually works?

[jade breach]

After 100 prompts, previous ones start disappearing, is there anyway to get them back without deleting recent ones?

[craggy current]

This! I didn't know this and would be great to know how to save stuff.

[icy abyss]

Does the app not work for others if you're using a VPN? Ive noticed when I have mine turned on the app isnt able to perform a search, it starts but never finishes.

[livid mantle]

im sorry but that would be too expensive

we need to maintain costs so ur prompts will dissapear

[autumn lark]

It's a visual bug.

Yeah unfortunately deleting the newest one is the only way. 😅

Oh well just file that info in for gg go next

[tight quartz]

email them. i got mine transferred to a new account but no data transfer atm

[frosty vessel]

why is reasoning only used on some queries?

[sleek topaz]

Is the latest sonnet 3.5 on phone app too

It still says latest fast model

[wraith crow]

yes it will also use the latest model on the phone apps

[heavy linden]

Anyone else here in a legal field utilizing AI?

[long flame]

Hi all, how many characters are currently available in the custom instruction input field for "spaces"?

[heavy linden]

Even though I set harsh instructions, perplexity still generates nonsense...

Inbe4 https://www.forbes.com/sites/mollybohannon/2023/06/08/lawyer-used-chatgpt-in-court-and-cited-fake-cases-a-judge-is-considering-sanctions/

[jade breach]

Gg?

[wind seal]

Hey
I am new just purchased pro
Can anyone share some resources and tips to get the most of perplexity ai
I am mostly in coding

[serene sand]

#💬│general message

[long hill]

is there an Discord Bot for my own server?

[wind seal]

Thanks

[livid mantle]

to maintain operation costs, sorry for the inconvenience.

[minor shard]

Yeah unfortunately AI cannot be trusted to be correct 24/7. one day maybe but right now we have to be the adults babysitting AI

[heavy linden]

True. I've tried almost every big gen-AI there is on the market currently and they're all creating their own jurisprudence. I simply thought Perplexity kept it real and based its answers on actual search attempts/findings. It doesn't.

[versed plaza]

One method to make things less worse is to babysit the AI through long series of instructions hand holding it through the system prompt

https://www.perplexity.ai/collections/advanced-ai-driven-fact-checki-a3cMcPR.QsKkCRZ79UKFLQ

The prompt I wrote makes it collect lots of varied data first, explain every step of its reasoning, give a grade 1/100 for its research and for how true it thinks something is, it also lastly decides how confident it is in its own answer.

The more that you make the AI think and rethink itself the better off you are against hallucinations

[heavy linden]

I went through the hassle to make a strict 'space' in order for Perplexity to do legal research, yet it still hallucinates. One thing I did not do is set temperature to 0 though, perhaps that would help.

[indigo valve]

What is the difference between this and scratchpad?

[versed plaza]

It was inspired by scratch pad

It's just tailored towards fact checking

[indigo valve]

And it's formatted nicer

[sacred crag]

Where is the proper place to submit a bug report for the new macOS app?

[minor shard]

Are you sharing this prompt? if not its ok 🙂 just curious as I love to play with prompts that people come up with

[versed plaza]

Oh really I have some really fun ones I've been looking for someone to test out

[signal hamlet]

Hey @sacred crag! Please check the #1140622008086970420 section to see if the issue has been reported, and submit a new one if it hasn't.

[versed plaza]

https://www.perplexity.ai/collections/advanced-ai-driven-fact-checki-a3cMcPR.QsKkCRZ79UKFLQ

This one the AI fact checker, it uses the same principle as the scratch pad.

https://www.perplexity.ai/collections/timetable-ai-powered-historica-YQPrR8SARiyT520Qo3b04g

This one is a special project I've been working on the last couple days. Go into it blind!

[strange tangle]

Hey, just joined the server, and I noticed during the onboarding, there's no option to choose mac app on where all I use perplexity from.

[versed plaza]

Let me know if you get any cool results from it! It's part of a bigger project I'm working on to make llm more believable

Technically the fact checker is too

Here's I guess a little how it works

The Concept:

TimeTable transforms the traditional historical Q&A into an immersive lunch date experience with historical figures. Instead of simply role-playing responses, the system creates a living, breathing historical moment complete with period-appropriate environments, supporting characters, and contextual details.

It works for more than just historical figures but anyone you can think of it creates a very, sometimes uncomfortably realistic RP chat

It also plays upon the scratch pad system as well, by doing extensive research before starting

[minor shard]

i can't see the prompt i dont think,unless im doing something wrong.

sorry 😦

[earnest kayak]

Hi everyone, I just sign up for Pro, which model most of you have it setup with? which one is the most recommend it? Thanks.

[full gyro]

A Pumpkin has appeared!

Time is up!

[harsh verge]

#1240356137497530408 message

[halcyon coral]

Hey @strange tangle!
Please check again

[compact bough]

Does perplexity pull youtube video transcripts as context?

[shadow granite]

Beta testing time

[sand sorrel]

Hii beta yess

[open furnace]

Hi

[vapid onyx]

I will say the MacOS app has been a tremendous QOL update to my experience with Perplexity

[rare wharf]

Agree

[viral current]

Shopping? Please - don't. Just don't.

[tame current]

Why does perplexity require access to the native speech recognition in iOS as well as raw access to the microphone?

[cinder comet]

https://www.theverge.com/2024/10/25/24279600/google-next-gemini-ai-model-openai-december

we will have gemini 2.0 soon

hopefully it will be added

[hazy moth]

I'd assume it's so it allows broader access to different speech to text AI models, since native covers apple models and raw covers other models like whisper.

[cinder comet]

While OpenAI CEO Sam Altman is doing a phased rollout of the successor to GPT-4, starting first with his business partners, my sources say that Google is planning to widely release the next version of Gemini at the outset. I’ve heard that the model isn’t showing the performance gains the Demis Hassabis-led team had hoped for, though I would still expect some interesting new capabilities. (The chatter I’m hearing in AI circles is that this trend is happening across companies developing leading, large models.)

[hazy moth]

I wonder if Gemini realtime is on the horizon?

Gemini live isn't a native speech model after all

[sleek vortex]

omniregressive

[hazy moth]

Capabilities in the AI landscape are increasing rapidly, it's just that they are scattered across different models from different companies

Realtime was a huge advancement imo

AI video too

Flux.1 as well

o1 and 4o-mini, as well as the qwen models. Gemma 2 2b, Llama 3.2, and more

Mistral is back as well, with large 2 and ministral

Google Illuminate was pretty awesome as well

I mean of course we aren't going to see the crazy gains with heavyweight LLMs specifically. GPT-4 set the benchmark pretty high at the start, and by then, OpenAI had been working a long time on LLMs that most people didn't hear about until ChatGPT appeared.

[wet lodge]

Where do I post for troubleshooting help? I’m a pro user

[serene sand]

https://x.com/aravsrinivas/status/1849879578124353774?s=61

Aravind Srinivas (@AravSrinivas) on X

@laqd99 Soon

Twitter · Oct 25, 2024, 6:23 PM

[tame current]

My custom UI, Before & After 😄 (bloatware visual clean)

[round plover]

pplx changing language?

[full gyro]

A Pumpkin has appeared!

Time is up!

[round plover]

russian, japanese, o.O

[near falcon]

Help! My perplexity just changed to korean

😭

but language options say it's set to English

[fierce sonnet]

We're looking into it! Thanks for flagging.

[supple shale]

lol just joined to say this

[vocal crypt]

Lmao just came here for the language issue

[supple shale]

super helpful tool but I can't understand the language now

[vocal crypt]

At least the app isn’t broken

[earnest saffron]

Mine is in spanish 💀

[supple shale]

bro this site is carrying me thru med school lol I need it to work

[mental sedge]

so the language issue isnt just me 😭

[earnest saffron]

fr i rely on this for ap psych

fixed now thx

[tranquil orchid]

Hey @earnest saffron!

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.

[harsh stag]

I don't mind if Perplexity integrate shopping. However, it should be a dedicated section alongside traditional generative AI answers, NOT trimming down the answers to 3-4 bullet points and shoving products in our faces 👎👎

[harsh stag]

it was my experience with "shopping" on perplexity a few days earlier when I changed my IP to the US, but things might be changed now, so please correct me if im wrong

[tame current]

true

[tame current]

Isn’t there some chrome extension that adds perplexity results to Google searches

Can’t seem to find it

Pls @ me if you know of one/if there’s one for perplexity

[marsh salmon]

a

[obsidian mantle]

Hi can I use perplexity here?

[serene sand]

https://www.perplexity.ai

[light barn]

There is this weird background when I open perplexity wiuth shift + command + p and the font doesn't feel Mac OS native?

[vernal warren]

I think it’s because it has two different ways of handling speech input. Cmd-Shift-V uses the raw microphone input directly, while Cmd-Shift-D uses Apple native dictation

[fathom blade]

I've been using the Mac App but reverted to the WEB APP because the FONT is too small, and I can't read it with 80 year old eyes. I use Perplexity ALL DAY, EVERY DAY, and love it, but the native Mac app needs to escape the "Fine Print" to be useful to me.

[fast meteor]

https://www.perplexity.ai/search/iphone-aFsBfrnFS3KX9.7U_OUcGA
the shopping thing dates back at least 5 months ago

[plain lintel]

hope im not too late for the pro beta

[drifting stone]

Is it just me or are the models revert to default/GPT despite selecting other AI models? Sonnet replies looks like GPT or default model replies (writing mode) 😔

[dense birch]

dude, why all model are defaulted to gpt4... gpt4 is nowadays have become dumber compare to other model...

[stiff monolith]

When will the pages and AI-generated images be available in the mobile version?

[minor shard]

Agreed. It needs to be its own tab

[livid mantle]

https://tenor.com/view/chost-machine-ai-type-typing-gif-481421031430735140

[serene fulcrum]

Reasoning beta is so cool

[cursive jacinth]

what sir? does it exist sir? most of us never saw that sir. You son of Aravind sir?

[cursive jacinth]

Thats their plan sir. Not OUR plan sir. 🤣

[sleek vortex]

😭

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@sleek vortex

[cursive jacinth]

seriously sir, without applying sarcasm sir, did you ever experience this so called reasoning beta sir? this new one sir.

[sleek vortex]

yes

[cursive jacinth]

You son of Aravind sir!

[sleek vortex]

😭

[minor shard]

lol u ask for no sarcasm and then you respond with sarcasm 😅

[sleek vortex]

just tell it to make a table

and it uses their reaosning beta

its alright

and only works for listing queries

[versed plaza]

Was there a recently an update where you can unattach a photo from a conversation finally? Cuz I feel like I got the option to do that with the clipboard but I don't know like it would be really cool to just have a picture important to one question

Also is it just me or are the voices more expressive than they were like before?

Mine seems to be doing things I don't remember hearing it do before

[dense kraken]

My Perplexity seems to be unable to access web search results. I confirm that I did NOT choose the "writing" option.

https://www.perplexity.ai/search/find-reddit-or-forum-articles-JPzYLTzIRDqsF9iROw13Tg

What's going on? Is it just me? Is it an outage? Or is it an intentional thing?

[torn sierra]

A few days ago, I came across this puzzle on Facebook. Naturally, my first thought was to feed it to ChatGPT 😉 Here’s what happened. Claude also can’t handle these matchstick puzzles. Sometimes it works and suggests moving a matchstick to create an inequality sign, but when you insist on keeping the equals sign, we get completely wrong answers, each attempt worse than the previous one 😉 Just shows how LLMs understand things 😉

[full gyro]

🎃 Halloween Pumpkin Collection Leaderboard 🎃

🥇 @minor shard (𝓙𝓪𝔀𝓷𝔃𝓮): 6 🎃

🥈 @sleek vortex (sneakyf1shy): 5 🎃

🥉 @serene sand (mares): 5 🎃

4. @last storm (kentavius): 3 🎃

5. @halcyon coral (IceLavaMan): 2 🎃

6. @rich sphinx (WhatIsAModel): 2 🎃

7. @native latch (humbled): 1 🎃

8. @brazen vine (Space_Army8485): 1 🎃

9. @wet quarry (CoreMT89): 1 🎃

10. @crisp zenith (ilikecheese): 1 🎃

Page 1/3

[harsh verge]

With a slightly more elaborate prompt. https://www.perplexity.ai/search/i-need-help-finding-detailed-i-3QhwupGcQ6qUqkDhkUIofQ

[dense kraken]

did you turn on pro or something

[harsh verge]

Pro Search, yes

[dense kraken]

Thanks for the help, but this doesn't address the issue of why my search doesn't use web results because you used pro while I didn't :-/
I suppose I'd try your prompt with not-pro search

Well, it still refused to search the web, despite giving me more detailed response

[opaque shoal]

Is the beta for shopping closed already? I clicked the link and it says “no healthy upstream.”

[potent rose]

I find the Finance product underwhelming - has anyone been able to use it for anything interesting?

Sure, the hard data is from Financial Modeling Prep, but just gives me basic stock quotes and some basical analysis tools.

[sleek vortex]

what would you like to see interms of finance features ?

[potent rose]

Asset correlations, constructing a portfolio, collation of reports with expected returns, fundamental analysis, etc.

[flint scarab]

why does this happen every time i open the perplexity mac app

[atomic stirrup]

ninjachat.ai is building something for this

[sand ether]

Does anybody know why I keep getting these pop-ups with the MacOS app?

[sterile mural]

nope

[dense ice]

i got perplexity for INFORMATION but lately when i type in items, now it's trying to sell me stuff. if i wanted to buy something, i'd go to google search. how do i disable pp from offering sale crap? or is this baked in?

[compact bough]

Does perplexity pull youtube video transcripts as context?

[alpine dove]

TLDR: "How have others been taking advantage of the new Space file upload feature?".... Long version: Sorry if this is in the wrong thread or has been discussed at length previously. I did look around first. I'm really optimistic about the new Space file upload capabilities. Although a couple of time I've had to remind it to reference it for every input without exception, I love that we now have more than 2000 characters for information settings / persona stuff or any of the 15 other synonyms. So far it seems like an improvement advantage. How else are people kicking the tires on file uploads?

[alpine dove]

I've gotten it to work a few times, but it is inconsistent. Even if you fashion your prompt explicitly and in detail. More often than not - it doesn't work. But that's one man's experience.

[compact bough]

I see, would add so much value

[west remnant]

Sonnet 3.5 is in the shared pro pool right? So like 600 queries or whatever

[worn zenith]

Yes

[ornate ember]

I am new here. Just curious in what kind of situations ppl use Perplexity and in what sitaution ppl use ChatGPT

[warped sequoia]

Hi, when will we see perplexity merch available?

[cyan socket]

guys is 40 o1 mini limit per day is good? i think its scam because its too little

[warped sequoia]

perplexity for everything = replacing google. havent used chatgpt for months

[full gyro]

A Pumpkin has appeared!

Time is up!

Collectors:
@warped sequoia

[raw depot]

WHAT'S GOING ON!!!

i've had this problem so many times now!
I did provide a PDF

[tropic beacon]

hey, is there a way to change mode for a existing converstation? so the thing is at first i use 'academic' mode where now i need 'math' mode but i don't see anywhere i can change it.

[light barn]

After I tried to copy and paste a text from VSCode?

[serene sand]

https://x.com/gunnargray/status/1850536770271580386

Gunnar Gray (@gunnargray) on X

Enter the observatory

Twitter · Oct 27, 2024, 1:55 PM

[placid coyote]

hi, is it just me, or some weeks back perplexity started "invalidating" longer opened tabs (maybe cloudflare?), so after going back, it reloads whole page and current position in the tab with pplx is lost (scrolled to the top)? it is pretty annoying

[granite vigil]

The new shopping feature is a surprise I wasn't expecting

I wonder what the end game is?

As a product owner I tend to get just as excited about speculating why features are released and guessing at the strategy as the features themselves

[fast meteor]

sometimes prompting it too hard will result in worse replies...

[summer halo]

Does anyone have the same issue? I select Claude sonnet 3.5 but the response is ChatGPT 4..

[stable notch]

here

Wait until 6 and he'll be Claude again, apparently he has a limit of 150 messages per day.

[west remnant]

Same here. I wish there was an actual indicator when you run out of sonnet instead of having to do an investigation 😑 I could sense the drop in quality

I didn't even know sonnet had a separate limit from the other pro models. Last I checked it ws 600 between all of them

(Except opus)

[summer halo]

Yeah… what happened? Wasn’t last time supposed to be 600?

[west remnant]

idk some people are saying it's a bug but maybe he's right about a new 150 limit 😭

[fading spire]

Made a mind map of the first chapter of the Bible comparing the timeline and information to backed up scientific historical evidence pretty amazing what perplexity can do. I did have to use the information generated into a different mindmap software cause asking it to make a mind map didn’t really end up to well.

[livid mantle]

Perplexity Space vs ChatGPT Canvas

[livid mantle]

what are your use caes for space canvas guys

[tame current]

Please for the love of god make it an option to open links from sources in default browser on iOS/android instead of the crappy in-app browser

[hallow bridge]

Is there no feedback channel for macOS app?

I can’t copy paste screenshots/images into the macOS app whereas I can via web. Have to manually attach image which is not ideal

[flint scarab]

so annoying

[tired topaz]

anyone having issues opening the MacOS app? Mine immediately crashes when trying to open.

[tame current]

None it’s useless

[livid mantle]

yes it is

i agree with you

its a marketing stratergy to distract us

to ease down our hype for gpt-5

not very clever you are sam altman

[vapid onyx]

I’ve forwarded this feedback on my end.

[worn zenith]

A big change with no prior notice? Where's all the extra money going since sonnet isn't nearly that expensive to warrant that

[worn zenith]

I'm saying it's a bug, at least they'd let us switch to less costly models like Sonar if true.

[wise edge]

➜ ~ curl 'https://docs.perplexity.ai/'
-H 'Upgrade-Insecure-Requests: 1'
-H 'User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36'
-H 'sec-ch-ua: "Brave";v="131", "Chromium";v="131", "Not_A Brand";v="24"'
-H 'sec-ch-ua-mobile: ?0'
-H 'sec-ch-ua-platform: "Windows"'

curl: (35) OpenSSL SSL_connect: SSL_ERROR_SYSCALL in connection to docs.perplexity.ai:443

docs breaken

[west remnant]

yea idk where he got that number. There was that upgrade to sonnet like 5 days ago but still... wouldn't be justified imo. Not at that number. Hopefully it is just a bug and gets fixed soon. I kinda like getting what I pay for

[worn zenith]

The upgrade didn't affect the pricing afaik

Also me too, I just resubbed and now met with this bug :/

[safe socket]

Hello everyone, quick question for you. Is using the "Sonar Online Large" LLM model a separate opion in Perplexity? Perhaps it's only available when you're using the paid option? If you can let me know, that'd be great. Thanks!

[full gyro]

A Pumpkin has appeared!

Time is up!

[slow glacier]

Hi guys, can't find API documentation. Is there API for perplexity

[versed plaza]

Thanks PPLX team for helping me wake up

The first site that I visit in years which has me actively learning rather than passively accepting

[summer halo]

All of the options are being run by ChatGPT… it is still not fixed on my end

[serene sand]

https://docs.perplexity.ai/home

#💗│sharing message ☕

[serene sand]

https://x.com/aravsrinivas/status/1850736323369554311?s=61

“Together, we form a necessary paradox; not a senseless contradiction.”

Frank Herbert, Dune

Aravind Srinivas (@AravSrinivas) on X

We’ll build and debug together the future of how people ask questions and get their answers and solve their problems

Twitter · Oct 28, 2024, 3:08 AM

[torn sierra]

Grok 2 "understands" images already. It doesn't get confused like Perplexity when another image is loaded in the thread 😉 https://x.com/grok/status/1850808322074509434?s=46

Grok (@grok) on X

eyes have been granted

image understanding now available

Twitter · Oct 28, 2024, 7:54 AM

[tender star]

https://www.perplexity.ai/collections/joe-maristela-bRRCVkgPQXO86KXaizZveg still no public view for shareable links for spaces/collections

[mortal talon]

Can’t we target anymore a part of answer to follow up in given context ? Feature similar to chatGPT’ one

[indigo valve]

You can't use any other model with free

[fast meteor]

apply for it and email api team about it to be on the (long) waiting list

[junior reef]

i've been on the wait list for a long time already and haven't gotten any updates

[fast meteor]

i guess it is case based review

[livid mantle]

@signal hamlet

@pallid elbow immidiately ping Alex

deepdive > #📙│rules

[pallid elbow]

@signal hamlet

[cinder comet]

lol

[livid mantle]

industry reports suggest that a new model, codenamed GPT-5, may be introduced by December 2024

[cinder comet]

who are you

[livid mantle]

GPT-5

you?

[livid mantle]

you tried to find me in search box

hahahahhahaa

[cinder comet]

one of your big fans