👉│feedback-general

[flat anvil]

But I don't want to put my home in my profile 😄

[sacred tide]

Great updates to the iOS app, but wanted to note that there is still quite a bit of lag, especially when asking more than two follow up questions (iOS 16.6, iPhone 13 Pro Max)

[visual marlin]

Would be nice to get inline latex 🙂

[sour acorn]

#🍏│feedback-ios

[hazy grove]

There are a bunch of channels that seem to default hidden for Discord members. Just setting the defaults to see all channels would fix so much of the confusion in here around reporting issues and bugs.

[rotund dome]

I want to be able to create folloup threads for myself. Now this is only possible when sharing the link of thread for external users. Love Perplexity new updates, I went PRO yday and cancelled ChatGPT Plus.

[fiery hinge]

I like Perplexity a lot and use it everyday!

Thanks for introducing Claude as an alternative

One suggestion / feature request that would help me. It would improve the search outcomes a lot if you used native language for the web search. Europe is pretty complicated. I'm Polish and live in Switzerland. I look up stuff related to Zurich (my hometown) and if I search in Polish almost nothing is found. In English the results are pretty limited too. German of course gives the best results.

[gilded creek]

Hi all I decided to try the premium features to test claude2 model in Perplexity. I gave it a shot, and while I'm impressed for the growing and dedication of Perplexity, it appears that claude.ai's chat outperforms perplexity witg the claude2 model when extracting insights from lengthy PDFs. 📚🧐 Could you share your insights on why this divergence in performance?

[bold blade]

I really don't see where the team put their "dedication" on. Maybe on something that I can't see and perceive...

the most significant change I noticed is the file upload, and then claude-2. But I don't even bother to use these twos on perplexity, cause claude.ai implements these so much better than perplexity themselves 😅

maybe the team is too busy on their way to fine-tune gpt-3.5, which I, personally as a paid user, get 0 value out of.

hey pplx team, can you please give us on what are the team's priorities after claude-2 and gpt-3.5 fine-tuned?

[long glen]

Curious if somebody else does also experience this problem with text overlap in Chrome addon code snippets. For me it almost constant.

[bold blade]

open the thread on another browser, if it still persists, then make a bug report

[visual bough]

Know issue, will be fixed 👍

[sour acorn]

https://discord.com/channels/1047197230748151888/1145348110474821722

[icy goblet]

Because it’s not any better than what was offered before? Is it not at least faster?

[bold blade]

it has always been bad for me

and now they've actually replace gpt4 with the thing they claimed "match Gpt-4 performance"

to be honest, I would rather describe what I want and how I want the model to response right in the prompt than being asked stupid questions by copilot

really doesn't help at all

[icy goblet]

You are seeing a meaningful degrade to the current gpt3.5 fine tuned vs gpt4?

[frosty nexus]

If you are talking about Copilot, at least for me it feels **worse **than before.

[icy goblet]

Got it. Thanks a lot!

[frosty nexus]

Maybe I just have go give it more tries. But since the latest update on Copilot I barely use it.

[ornate cipher]

Друзья! Такой вопрос! Я оплачиваю 20 долларов за chatGPT, 20 долларов за Notion, и с недавнего времени 20 долларов Perplexity. Что интересно - все эти сервисы пользуются одним и тем же ресурсом - GPT-4. Вот было бы хорошо если б у вас была возможность ввести ключ API от моей подписки chatGPT и платить чуточку меньше. Я не говорю что не хочу чтоб вы зарабатывали, но можно же сделать чтоб для уменьшения ежемесячной платы люди приходили со своим "движком".

И у меня еще один вопрос, простите если перегружаю: бы ло бы классно, если-б у вас была семейная подписка или аккаунт с мультивходом. Я не про тот мультивход, где несколько устройств с одним гугл аккаунтом, а несколько устройств с разными гугл аккаунтами. Пусть их будет 2 или 3 - чтоб я мог дать возможность пользоваться более качественным поиском на движке GPT-4 своему сыну и супруге. Заранее благодарю за ответ. Мне даже кажется, что всё это есть, но я просто не разобрался

И, простите, еще: НЕ нашел хорошую статью про исследование качества работы поиска в Perplexity на разных движках. Сама Perplexity, удивительно, похоже о себе ничего не знает. !!! 🙂 Может кто-то из сообщества даст мне ссылку на хорошее исследование, сравнение, академического характера, не рекламное с общими словами. Наверняка такое существует! Уфф... Простите, загрузил :-). Всем хорошего дня

[median yacht]

kept having this problem in Chrome, any one know why?

[frosty nexus]

Hey, are you using Perplexity's AI Companion extension?

If so, just disable / uninstall it for the moment. The team is aware of this issue.

Had that happening a week ago.

[frosty nexus]

Let me know if it worked!

[fringe pawn]

I get much better answers on phind.com for programming questions (havent tried much else in phind). And that is for their gpt 3.5 free version. The same in perplexity can be frustrating as it won't understand questions or repeat answers sometimes.

[sour acorn]

even though I am fan of Perplexity, I can vouch for this comment. I mostly tend to go towards Phind when that happens.

[sudden timber]

im having this issue since this latest claude 2 update. I couldnt access Perplexity Ai without using VPN. I didnt use any perplexity chrome extension.

p/s: btw im using Perplexity Pro from Malaysia

[visual bough]

Have you already tried to verify on labs.perplexity.ai?

[sudden timber]

ive tried using labs.perplexity.ai but STILL having the same issue

[visual bough]

Then it will be due to the IP, it is best to send your IP to a dev so that it can be checked.

[median yacht]

it is solved, it is indeed the AI companion extention that caused the problem, thanks!

[dusty bison]

Is there API access for developers?

[glacial granite]

Hey @dusty bison! Perplexity AI does not currently provide an API.

[south ether]

Hello, can you send me your IP address through DM so we can fix? thanks!

[sudden timber]

sure. thanks

[old tartan]

OK, how do I get Perplexity to actually do the thing that I ask it to do. Are there certain prompt keywords I should be using? Because right now when I tell it to do XYZ, it sends me seven sources on how to do XYZ. I understand as a learning tool that is very helpful. But i’m not asking it to teach me how to do XYZ. I want it to actually do the thing that I ask it to do. Even with copilot this is becoming a problem. I have Google search, Google search is perfectly fine. It’s beyond fine. It has decades of data from which to draw from. If I wanted to search for how to do something, I go there. for free. But I’m paying $20 a month so that Perplexity can help do all the mundane things that I waste my time with so that I could do other things. But now I’m wasting my time trying to get Perplexity to actually do the thing that I ask you to do. So I am assuming that I’m stating the prompt incorrectly, or that there are some trigger words that I’m not using. So if anyone has any guidance in that regard, please let me know. Because right now I’m basically paying 20 bucks for, free Google search. And considering does the exact same thing that Perplexity does, I don’t need to pay $20 for it if it’s not going to actually do the task that I’m prompting it to do.

[spice furnace]

I think I'm going to need a default focus. And Claude always tells me that it is summarizing is unsettling UX. I love to summarize with Claude but that's only about 1 out of every couple hundred interactions. Somehow it does the right thing but I'm always rready for it to smash my content rather than evolving it.

[spice furnace]

It seems to get around to it but I have encountered that a few times too. Are you using Claude? It seems to need a bit more prodding and cajoling.

[covert sun]

Hey @old tartan! Could you share a permalink of that result. We need to see it, to debug or suggest some improvements. I suspect you are trying some generative/processing task, but using the All mode, where you could use the Writing mode. But of course, the best way to solve it, would be to see what you are actually trying to do.

[pulsar ravine]

This has probably been mentioned before but I'd like to see a way to organize my threads into folders or groups. On top of this, nesting folders would be even more cool.

I use perplexity in combination with Obsidian (PKM tool). Many times I save threads from perplexity into obsidian because I want things organized, and then I delete the thread. This is a bit of a pain.

Note: I posted this both here and in #💭│ideas for obvious reasons.

[frosty nexus]

This is apparently coming soon: https://discord.com/channels/1047197230748151888/1130050097800028201

[pulsar ravine]

Unknown channel / Link. Is it the pro lounge?

[visual bough]

Yeah it´s the pro forum.

[south ether]

coming soon! 🔮

[flat anvil]

Did you try to set the focus on writing? That was my best experience for the most tasks.

[cedar crystal]

Would love to be able to copy the entire thread 😉 this would simplify y workflow.

Also, it would be amazing if there would be an option for the citations to be anchored on words from the write-ups, not on numbers.

[vocal depot]

Hello, I am having trouble accessing perplexity on google chrome, any idea why I may be having such problem? it works perfectly on safari though.

[covert sun]

Hey @vocal depot! Do you mean you can't pass the captcha?

[vocal depot]

sort of, yes. what I mean is that 'it' doesn't get to the captcha. @covert sun

[covert sun]

Got it, @vocal depot, please try to get to it and solve it at https://labs.perplexity.ai
and let me know if the main site works well after that.

[vocal depot]

it worked! thank you. @covert sun

[west turtle]

Yes, the responses are noticeably worse now. A few of my conversations degenerated into gibberish.

[bold blade]

"degenerated into gibberish" - wise words usage there 👏🏻

[queen knot]

Hi @covert sun
IceLavaman asked me to contact you

I am a security researcher and I would like to report a security bug

Where can I do that?

[heady plover]

I have noticed with joy that now the Android app answered a question in German also in German. The web version still answers in English though, even when the question is in German. It would be great if the web version would also answer back in the language of the question.
I know that I can set the language in the settings, but I am often changing language so it's great if Perplexity would just answer automatically in the language of the question.

[frail sleet]

https://labs.perplexity.ai is not working can someone please fix the issue.

[weary patrol]

It's working for me. What's the issue? I'm not a developer tho.

[sour acorn]

#🦙│feedback-playground

[turbid wasp]

It's annoying that when I give feedback about a result, it just gives the same result with an apology at the start

[broken juniper]

Hello, I'm trying to pull the list for House of Representatives for North Carolina (federal positions). The response I'm receiving is the prior Congress (117th Congress), not the current 118th Congress. I tried to introduce internet links and prompt corrections; however, I'm still getting the wrong response. My thought: North Carolina had one additional federal Representative in Congress so the map was redrawn. Perhaps, this is an underlying reason the database is not updating. This is just a guess => I need help in getting this corrected. Side note: this is my first message, apologies if this isn't the right thread.

[hollow hollow]

I tried asking what was the highest unemployment rate in the past 80 years in USA? and it gives me answer in 1933, which is 90years ago https://www.perplexity.ai/search/8d4a6435-4262-4373-942a-6ea7407afc33?s=c

[queen crane]

I use Perplexity.ai almost daily. In the last week or so the search results have deteriorated significantly. It hallucinates really extremely. Out of six links, five of them are completely unresponsive

[knotty lagoon]

Hi there. I think there is a pro channel somewhere in Discord? Or at least there was at one point. I am a pro subscribed but I no longer have access to the channel. I checked the Channels & Roles section but couldn't find anything about this there. Anyone have any guidance on this?

[visual bough]

You can leave the server and rejoin via the Pro invite link in your settings

[queen knot]

Kindly check my DM @covert sun

[hazy grove]

Really wish these things were prioritized. Would prefer to spend on Perplexity Pro but these are dealbreakers.

[covert sun]

Hey @hazy grove! The issue with the username was fixed. Please check this thread: #1145171019523424336 message
Let us know if you are still facing any issues.

[hazy grove]

It worked, thank you!

Will give Pro a shot 🤝

[icy goblet]

Hey

Please add the function that Claude.ai has on automatically pasting a large amount of text as a file directly

This is an incredible feature that perplexity is missing

Turning a big amount of text into a pdf before putting it in perplexity is too much of a hassle

[snow mirage]

Did you get an answer to this? I've noticed similar, except replies are telling mehow to find my answers, rather than scraping the answers like it used to.

For example real estate in certain geolocations with specific parameters. It used to cull a list. When I use Bard, it gives me a list of properties.

[snow mirage]

Have you tried pasting a weblink to the PDF? That seems to work for me, or it did a week ago. Granted that means you have to upload your PDF online. Maybe try some other publicly accessible PDF?

[icy goblet]

I am not interested in web links though

Sometimes I just have a very long transcript or something like that that I need to get help from Claude for a summary

[visual bough]

This has been suggested many times, now the devs have to decide if and how they want to integrate this function.

[fair helm]

Perplexity Google Chrome extension is interfering with Cloudflare causing infinite loop of checking if connection is secure. I have been trying to solve this for hours. I disabled Perplexity extension and it worked.

[crimson nymph]

It's a known issue. If you want to keep the extension enabled you could also visit https://labs.perplexity.ai/ solve the captcha there and go back to https://perplexity.ai/

[eager cloud]

Why ios app can't use focus? When I need to avoid search I have to use web broswer.

[visual bough]

This is in progress and will be available soon.

[south ether]

we are aware of this issue, we would advise users that experience CF to disable the chrome extension. if anyone is experiencing this issue and does not have the extension installed please let me know, these issues will be fixed but it's the best fix for now.

[midnight orchid]

it is annoyint that the maximum query is only 2000 characters, especially when i subscribed for the 32k context and the 100k offered from Claude, also, isn't it strange thta it isnt' compatible with Docx files?

[crimson nymph]

it is 2k tokens per query not characters. And docx support is coming in the future. It's not that strange since docx files aren't that easily readable as txt files.

[spice furnace]

We are back in the day like when computers had 4 and 8k. Memory was precious. My last docx file was ten times the size of the pdf and the pdf was 30 times the size of my text file. 7 megs/724kb/24kb

[ivory elbow]

I'd like to be able to export perplexity threads as markdown files

[echo jackal]

what advantage does the paid plan for perplexity $20 have over chatGPT plus with the plugin metaphor search, and there are other search tools out there ie andi search, gigabrain, GPT GO, you, etc for free.

[echo jackal]

has the perplexity team thought about making other AI tools for coding? Or adding those tools to the perplexity paid version? I mostly use AI for coding tbh lol

like issue with coding noadays with AI is it can't handle more then like 200 line programs without running out of tokens and getting confused

i can't wait until it can do like 1000 line programs

[limpid sparrow]

Is gpt go good?

[echo jackal]

its ok, it kind of gives answers like koala, one of the best ones i have used that searches the internet they changed to a paid plan was totalquery plugin on chatGPT , sucks i can't use it now because you only get 15 searches a month where it use to be unlmited, but that was my go to

now honestly i mostly use metaphor search plugin on chatGPT plus

[wispy flume]

On the other hand, you have only 50 available messages per 3 hours in chatGPT-4. Is it enough for you?

[echo jackal]

when i'm using it for programming, it has not been, and i have had to either wait or buy poe where i get like 600 or something a month

i have not seen that message on chatGPT latley did they change that rule or just remove that message? lol

[wispy flume]

But you can use perplexity AI pro writing mode instead of Poe - it has 600 per day. In chatgpt this constraint has been a long time ago

[flat anvil]

https://www.perplexity.ai/search/7fdad5a3-aec9-4ed4-bb8d-b63899d834ef?s=u

Please scroll down in the thread.
Perplexity doesn't know Adobe Firefly AI.

https://www.adobe.com/sensei/generative-ai/firefly.html

@south ether is the perplexity AI not updated with 2023 news?

[crimson nymph]

You are in writing mode. That does not have internet access

The cutoff date for gpt is sept 2021

[flat anvil]

Ah okay, yeah I prefer to use writing, because the AI does speak more "humanlike" 🙂

[quartz dagger]

wait you talking about programming and AI support and don´t even mention phind? That sounds strange, I wasn´t aware there is much around that comes anything near to what phind is able to, it has a reason its called "search engine for developers"

[echo jackal]

well that's why i go on discord to learn about things like this thanks 🙂

[sour acorn]

For Real

[quartz dagger]

haha great glad I could help here for one time, normally I am on the other side where I need help instead of being able give some lol

[quartz dagger]

fr fr 😉

[echo jackal]

yeah, but i have not gotten those messages recently at all on ChatGPT plus

yall try out feedly? feedly.com?

kind of a nice news aggregator

[clever brook]

the perplexity extension tends to translate everything when the chosen language is different than english. I looked up the lyrics of a song and the pplx extension gave me the translated lyrics instead of the original english ones.

[heady plover]

I use inoreader

[heady plover]

the only one that even comes close to perplexity (3.5, not gpt4) is fastgpt, all others are horrible except maybe phind

the main advantage of perplexity is the low amount of hallucinations and gpt4

also, copilot is useful

[marble patio]

Not sure if this has been addressed, but a lot of times when I navigate to perplexity.ai, I'll get the Cloudflare prove-you're-not-a-bot page. This is fine except that a lot of the time it falls into an infinite loop of me checking the box, Cloudflare starts to load, then the box reappears. Clearing cookies doesn't always work.

[sour acorn]

#1149876534325288981 message

[icy goblet]

Please make it easier to switch among perplexity, GPT-4, and Claude 2

[sour acorn]

#1145894530239975474 Dk when its coming though

[umbral tangle]

iOS app feedback based on my frequent inadvertent back tap when typing in new thread -

When creating a new thread, it would be beneficial to preserve the user’s inputted text, even if they inadvertently navigate away. Upon returning, the saved text should be highlighted by default, offering an opportunity to easily delete or modify the content, thereby optimizing the user experience and minimizing time loss.

[severe nebula]

I have been doing a bit of reading. The FAQ section is pretty good, but I feel you dodged the question on the two points in the screengrab. Nowhere do you state what part of our AI search history is stored. You do address PII well though. Do you think you can update this to cover the points?

You do mention further in that Search queries and feedback are considered in AI learning

[heady plover]

i really want to thank all developers for releasing collections!!! i honestly needed it for quite some time now, cause my list of threads got so big that i needed to search in my list for the thread that i want to use! KUDOS! you are waay better than openAI (plot twist, they did not release collections so yeah)

[visual bough]

With this you can sort your threads into individual categories and also create threads together with your friends in one category.

[icy goblet]

May I suggest setting a channel for Claude 2, one for GPT-4, and one for Perplexity? (For the different models available on Perplexity)

I would think I am not the only one who has feedback on the Claude 2 implementation in Perplexity

[clever brook]

whenever I click on the copilot round arrow to regenerate an answer, it exclusively answers in english and not in the language set by the user.

[visual bough]

Have you switched on the AI Profile?

[clever brook]

yes

ill try again

yup does it again

[visual bough]

I can't reproduce the problem, can you please take a screenshot of the AI Profile page and send a link to an affected thread?

[clever brook]

which thread in particular?

[visual bough]

A thread where the answer is in the wrong language.

[clever brook]

oops i meant which discord channel do i attach the thread?

here?

https://www.perplexity.ai/search/1bb22666-9828-4f81-9180-573c27f1ff8a

here is the thread in question

[visual bough]

Thanks and the Profile is active?

[clever brook]

yes it says pause as well

[visual bough]

And it only happens when a response needs to be regenerated?

[clever brook]

yes only when i click on the copilot round arrow at the bottom left of an answer

this one

[visual bough]

I see thanks

[clever brook]

the original thread in french was this one https://www.perplexity.ai/search/b33b1524-41ab-43af-9aa5-c0317a4fa558

[visual bough]

Would it be ok for you to share your intro from the AI Profile? (DM also possible)

[clever brook]

sure dm

[nocturne valley]

I would really suggest that we could have a button that can delete all threads at once, will be very helpful.

[livid goblet]

Started using perplexity today on web!

Couple of suggestions:

Pain point: sidebar expand/collapse button are in different places! Would be nice to have it in one place for easy toggle.
There should be a feedback button accessible no matter where in the UI you're at, for less friction in submitting feedback.

UX suggestion: devise a fun to watch animation for text that's being generated. A simple opacity change in the text seems less satisfying compared to how chatgpt and bard do it!

[chilly turret]

At https://alpha.perplexity.ai there is an option to delete all threads. Probably this function will arrive in the regular version aswell.

[gray pumice]

direct result of not allowing this to happen is broken image

hopefully something can be done about this

I'm not using any VPN tho

[glossy heath]

how did I use all my copilot uses today when I only searched maybe 20-30 times?

It told me I had 580+ remaining 10 minutes ago...?

anyone else report this problem?

[visual bough]

The 600 uses refer to Copilot, GPT-4 and Claude 2.

[glossy heath]

weird. it was a google chrome issue. I logged in using incognito and now it shows 508.

@visual bough yeah so 1 search is 1 use. Right? Want to make sure there's nothing tricky going on with how it calculates '1' use

very good. thank you!

[barren scarab]

There's some issue looking at my threads in the website through Firefox. Either keeps loading or says there is nothing there. Anybody else facing the issue?

[visual bough]

Currently there are overloads/bugs, please do not open bug reports and please be patient.
Perplexity is stable(er) at https://alpha.perplexity.ai.
Thank you!

[agile ember]

any plans to develop a plugin in VS code ?
thanks!

[potent escarp]

thank you for this! alpha works just fine!

[left epoch]

What happened to threads within the main perplexity page? All of mine are gone

[crimson nymph]

#1151514486008717352 message

[left epoch]

Not applicable since I was NOT using Alpha, but I've asked if we HAVE to use Alpha to have access to our past threads (tho that seems counter intuitive). Thank you for the link to that.

[crimson nymph]

Some users are currently not able to access their threads on the main website. You can still access them on the alpha website.

[sinful stone]

Won't it be much cooler to have Create and Add to Collection button when adding a thread to collection?
RIght now:

• click on three dots next to thread.
• click Add to Collection
• click New Collection, and Create
  Now again click three dots next to thread and Add to Collection and select the new collection.

Bug: I cannot choose emoji after selecting the skin color for Collections.

[karmic bridge]

why my card is refused suddenly? I have tried diffferent credit cards, they were all refused. But I used to pay by the cards, why it is suddenly not working?

[visual bough]

Hey @karmic bridge!
The best thing to do is to contact support for payment problems -> [e-mail]

[karmic bridge]

Thank you！

[odd dragon]

can Perplexity replace google.com ?

[heady plover]

for answers sure, but for exploration, and discovering new things etc. I find traditional search engines to be better since perplexity doesn't have as many sources

[pastel ruin]

The app UI should definitely be more like the website UI, it's very annoying to navigate

idk why popular threads are above the actual threads you'd want to use

[sinful stone]

Please make Collections available on iOS app!!!

[humble spear]

On the website, when editing a query, sometimes I scroll all the way up to the query before I remember that the edit button is below the answer. Not sure what UI to suggest around this, but maybe it is as simple as putting an edit button somewhere near the query itself.

[rich barn]

Why is the generation speed of Chinese responses much slower than English? It may take up to half a minute to generate a conversation

[surreal pine]

5 mins

[visual bough]

Known bug:

1. go to https://labs.perplexity.ai
2. solve the captcha
3. go back to https://perplexity.ai/

[heady plover]

I just asked 0x503-0x40 once before, and then as long as I ask directly without adding anything, it will directly return 0x503-0x40, but what I asked was 0x503c-0x40.

Is this a bug?

Shouldn't it detect what I type and then answer it based on the question I asked?

It answers the question I asked last time

[drifting nymph]

That accuracy... How can it missspell Anthropic as Anthropoc while it was correctly spelled in the question? https://www.perplexity.ai/search/b43abb1d-ee6d-4e37-b20c-00aaa1a49ab9?s=u

[visual bough]

This can happen, but you can report the answer with the flag below the answer.

[surreal pine]

ppx is halluncinating at new level on newer questions

[visual bough]

You are also welcome to report wrong answers with the flag 🙏

[heady plover]

HI everybody, I love to use ppx in academic mode. Today is the first time I saw it hallucinating. It made up every article / link it provided. The first link exist, but talks about a totally different topic. https://www.perplexity.ai/search/Show-me-research-D_xjspi3SpOQCPhCkBEIIA?s=c

[crimson nymph]

looks like a bug. It actually didnt use any sources and seems like it’s in offline mode.

[covert sun]

Hey @heady plover! Spezi is right, I see there were no sources attached to this answer. I tried to repro, but I do get the sources and proper answers now. Let us know if you're still facing this in a new thread. We'll investigate anyway.

[crimson nymph]

I can only reproduce it with the perplexity model selected

gpt4/ claude seems to work fine

Perplexity model: https://www.perplexity.ai/search/Show-me-research-fwvhtG8mQD6slTpSDbxctA?s=c

perplexity model with extra ? added to the question: https://www.perplexity.ai/search/Show-me-research-xZDEK4hGTyCFVrZ9hZyHZg?s=c

seems to be something wrong with the question.. when I add an extra “?” its working fine @covert sun

[pulsar ravine]

https://www.perplexity.ai/search/c9a51821-ffce-4a8a-b1cd-2f24b0b312bb?s=u
The last question of the threadd. It doesn't seem to give me the answer I'm looking for, probably due to lack of research on this.

[heady plover]

Thanks! You are right. Now, I realize it answered like in offline mode.

[turbid wasp]

Hi! It's annoying that when I give feedback on an answer that's incorrect or sexist or whatever, it just says sorry then exactly repeats the bad answer again.

[narrow forum]

Can start a new thread with instruction to not to say sorry

[junior pasture]

This is a specific issue related to model testing. In theory, GPT4 should be able to answer correctly, but GPT4 with pplx did not answer correctly.

The correct answer is to consider 4g and 5g as quality concepts; the incorrect answer is to consider them as communication technology vocabulary.

[heady plover]

is that telegram ?

[junior pasture]

Yes, communicate through Telegram bot and ChatGPT API.

[narrow forum]

Did you test with english? Maybe chinese is less accurate

https://youtu.be/7iU6K7NccXk?si=pofH5SAay3152Whf

In the video he explains in detail how Perplexity works

[agile ember]

what a fun question. people these don't have such Chinese literature standard.

[agile ember]

does yiyan works for this command by the way?

[elder barn]

How is perplexity a competitor of Google lol

[frail fossil]

When I need information on something, it’s easier to find via perplexity vs googling, since perplexity will summarize the info for you. Google has something similar, but it isn’t as comprehensive and it sometimes doesn’t work.

[agile ember]

theoretically, yes you are right
for example to construct a table comparing iphone models.
if only perplexity is working accurately.

when we use google for this information
we know to click on GSMArena.

attempt to improve my prompt
by checking the website GSM arena
construct a table for comparing iphone 15 14 13 pro max with respect to
cpu, ram , screen size, price

public:
https://www.perplexity.ai/search/e53fa73e-97c0-41b3-8ad4-aeb8ffbb3d75?s=u

[elder barn]

I am not talking about thet. My point was it uses Google search api. How can it be Googles competitor when it uses Google search under the hood lol

[frail fossil]

I don’t need to go to gsmarena. Perplexity AI: Please construct a table comparing the available iPhones with respect to
cpu, ram , screen size, price, including the new iPhone https://www.perplexity.ai/search/Please-construct-a-c0Gs2eiQQi.YxvXEJlBfhA?s=mn

[agile ember]

am i paying a fake product? when i try your query on GPT-4
it is even worse
https://www.perplexity.ai/search/1038634b-1578-4fac-bc17-d16ec37ca368?s=u

I have turned off every customization.
no personal profile

[visual bough]

What do you have in your AI Profile?

Ohh

[frail fossil]

Oh I see. I was using copilot

[agile ember]

good catch
same query with co-pilot still not working as expected for me
is a bit better, still not very acceptable
https://www.perplexity.ai/search/7a09c1cf-a18f-40f7-83a6-248039d3eb59?s=u

did not activate profile

[frail fossil]

Weird. Same on my end

On the profile not being activated

🤷‍♂️

[agile ember]

but thanks for your sharing which at least let me know some of the peers are seeming using another backend ?

or let them know there is a large discrepancies across user, even if paid.

[dusk aspen]

The variability in responses is probably due to ambiguity regarding what exactly the AI should search for when considering "available iphones", it therefore probably proceeds along a different search pathway each time, thus considering different sources with variably up-to-date information. You'd be better off, specifying each iPhone model you wish to compare explicitly.

[agile ember]

1. the problem is not about which iphone model it gives, it's about they have no infomation about the CPU and incorrect ram was included in the table
2. previous threads in pro lounge has talked about specific phone model search. the table is still giving false information. FYI https://www.perplexity.ai/search/please-summarize-a-qtIAuH9oRYaxC9H2fFVuqQ?s=c

including false information is way worst then leaving it blank because if i am to explore new topics, i can't judge its correctness just like iphone spec tables.

[dusk aspen]

Check out this response to my query in this regard.
https://www.perplexity.ai/search/Please-construct-a-zMp6XKfTQO.ZhJuWJ8E2hg?s=c

[agile ember]

alright, it seems they have fixed the issue by referring to apple official website.
appreciate their quick patches.
now I got the same result
https://www.perplexity.ai/search/1a7315bf-24d2-4a61-99c5-45eb62c9261c?s=u

seems they really take users feedback.

[sinful stone]

Why the related section suggests answers instead of query prompts?

[visual bough]

This is a known bug, in the future this function should be improved.

[heady plover]

I wonder if a recent update could be added to this page so that users know what has been updated. I think this is a good way to let users know about new features.

It’s not just after using it that I discovered, oh, there are new functions.

Just like Google bard, you can view the change log

[sinful stone]

Ok.

[twin siren]

Am i the only one feeling it or perplexity ai quality has dropped significantly in the last few weeks.

[narrow forum]

Or your prompts are not as specific

[tame adder]

Hey, I've noticed there is no option for 2-Factor Authentication (#2FA). Is this going to be implemented anytime soon? Obvious data/privacy concerns. Cheers! 🤓

[bronze thunder]

Yes, I feel it too. The AI quality is dropped and it's not worth my money for the pro version

[chrome rune]

In the desktop browser, if the response has images, allow left and right arrow keys to view each image (similar to other sites)

[crimson nymph]

you can report wrong answers directly on the website by clicking the flag icon. this will help the team improve the answers

[vapid hare]

what is perplexity

Quick Search
Answer
Perplexity is a term used in information theory to measure how well a probability distribution or probability model predicts a sample1
2
. It is a way to compare probability models, and a low perplexity indicates that the probability distribution is good at predicting the sample1
. In the context of Natural Language Processing, perplexity is one way to evaluate language models2
. Perplexity is defined as the exponentiation of the entropy, which is a measure of the expected, or "average", number of bits required to encode the outcome of the random variable, using a theoretical optimal variable-length code1
. Perplexity AI is also the name of an AI assistant that helps with research by providing suggestions and sources in response to queries4
.

i was talk about perplexity but it forgot after one conservation

[visual bough]

Just to understand: You started a new thread and asked Perplexity about something from the previous thread?

[vapid hare]

no i was asking from same tread

[visual bough]

Please share a link to the thread.

[vapid hare]

how to shared i just saw it from discord and went to see

and not login

[visual bough]

I don't understand, please send a screenshot from the location you see the answer from Perplexity.

[vapid hare]

it's gone i was search discord i found perplexity and went to see what is it then i send the feedback i closed the tab and the tread it gone cause i didn't login

[visual bough]

Without a specific link I can't help you, do you perhaps still have it in your history?

[barren scarab]

When flagging an answer, does your team need only the right answer as it's shown. Feedback and describing what went wrong is much easier to report I feel.

[clear token]

Does Claude 2 use more sources than gpt4

[silver iron]

Why does pplx img gen generate such bad images? What model do you use? Can't they use a better one as a midjourney? The images are prettier

[dire pewter]

Is it not possible to modify the AI profile directly from iOS app? I can't seem to find the option

[visual bough]

Hey, @dire pewter not possible atm, but will be added in the future!

[placid aurora]

i'm not sure if it's normal but when i turn on copilot, it asks me what I asked in my question. it just asked me back.

Things like this, happened to me multiple times.

[icy goblet]

Which GPT-4 model do we get on Perplexity?

[slim dawn]

There are errors when we request lyrics for songs for example: "what are the lyrics of drake only you freestyle?"

Needs to be fixed or worked on 🙂

Thanks for such a great service tho!

[dusk aspen]

Quite a few times Copilot acts dense like this, just ignore it and skip

[rich barn]

Before the Chinese question will be much slower than English, now the Chinese question directly does not give the answer, the answer part is empty.

[visual bough]

Can you please share some threads (link)?

[rich barn]

https://www.perplexity.ai/search/1e3d8753-6ae6-46b5-877f-f6d643385d04

[rich barn]

ok
It looks like English will do the same

[rich barn]

I saw answers on the shared page, but not in my iOS app

weird

[calm imp]

Hi, I am new to this forum. I was wonderingn whether it would be possible to put a selector button for the desired language model (Perplexity: GPT3.5, Claude, or GPT4) on hte main page and not require the user to set this in settings? I switch back and forth depending on what i am doing.

[visual bough]

Hi, yeah suggested many times. The devs are definitely working on improvements!

[mint wren]

I just wanted to say that Perplexity is a banger, it writes the best emails, I mainly use Perplexity for this

[calm imp]

@visual bough: cool name. Thanks for the heads up.

[wispy flume]

I've got the same one

[keen relic]

Hi, I run two different browsers on desktop windows 10 and access Perplexity from two of the browsers. Within the last two or three days I've had a problem with Chrome passing the cloudflare security connection test. I disable one plugin for DuckDuckGo Privacy Essentials and clear cookies, but probably not cache, and closed and started Chrome, but still not getting passed Cloudflare secure connection, I am not using a vpn or proxy site, any ideas? should I log a bug report? the thing is I have no connection issue using Firefox, don't even get the Cloudflare security check message. Same desktop, same time, same internet access method, no vpn, no proxy.

[crimson nymph]

do you have the perplexity chrome extension installed? If you have please disable that extension and try again.

[sage vine]

I don't know if you (the developers) have any control over this, but it drives me crazy that matrices are written as [0 1; 0 0] for example. I know that it can display latex because it did it before already, so displaying "\begin{pmatrix} 0 & 1 \ 0 & 0 \end{pmatrix} " (of course the output should look like in the picture provided) instead as well would be great.

[keen relic]

Yep, that was it, thanks.

[tough halo]

Hello everyone,

• Is it possible to make the threads added into a collection disappear from the main threads page? If not, do you have any plans to add this as a feature?
• Can we have sub collections created inside collections? If not, when do you suspect this would be added?

Thank you in advance for your response,

[surreal pine]

I find this answer partial correct

[barren scarab]

You have an option to flag the answer at the bottom of the response.

[unkempt rivet]

math error

[tough halo]

Is there any chance I could get a reply to the questions above from Perplexity's support?

[visual bough]

Hey @tough halo,
these two functions are not (yet) available, but interesting suggestions!

Collections are also only recently available and this feature will still get some updates.
I see your suggestion already has 3 supporters, the devs will certainly take a look at it! 👀

[gaunt haven]

Claude does not finish the entire text.

[gray pumice]

https://www.perplexity.ai/search/e5da1c3c-e7ed-4873-b287-431d4d60bc14?s=m

Welp...

[dusk aspen]

Yeah, use copilot here for stuff like this.

[gray pumice]

so you basically say : pay for correct answer?

I am actually reporting this as incorrect answer so they could improve it.

Also, I actually tested this because man said on youtube video : I asked ChatGPT this question and got this reply

and Perplexity's answer is apparently not different - logic is wrong

Also, is this the world we ought to live in now?

where if you dont pay for correct answer you can just get incorrect reply and live with this.

so truth is paid for and not free?

[gray pumice]

near the end of the movie Tar Lydia Tar is watching video on tv of conductor giving a speach about music. who was that conductor?

current answer I get when I ask: which conductor is lydia tar watching on tv tapes?
Therefore, Lydia Tár is not watching any conductor on TV tapes, as she is a fictional character in a movie.

which is not true

Apparently with specificity it finds it: https://www.perplexity.ai/search/Who-is-the-mn5m_.oXTHq.B2LrjTpFHQ?s=mn

• thanks Spezi

But how specific should we be to get right answer

[visual bough]

To be honest, such errors are quite normal.
GPT-3.5 simply does not have a mature "memory" to understand such logical things.
But if you adjust your prompt, for example by adding that a towel needs 3 hours to dry, you will get the right answer.
The developers of Perplexity can't do much about this, the only thing you can do is to use a more intelligent LLM, GPT-4.

[heady plover]

I personally think reporting it is fine, if perplexity were to ever finetune a model, it would be helpful for them. but also, gpt 4 costs more than 3.5 so they can't give it away for free

there are 5 free copilot uses per 4 hours in the free version

[dusk aspen]

GPT 3.5 is notoriously bad specifically for questions like this, so its unlikely its going to rectify itself, even with sources, when it comes to mathematical reasoning, that's all I'm saying. There are limits to what the perplexity devs can do since the underlying model is still from OpenAI. On a more positive note, it seems the recent release of GPT 3.5 instruct by OpenAI might remedy this since it hasn't been lobotomised(fine-tuned) specifically for chat-related functionality and verbal niceties, so it will probably be much better at reasoning. I'm looking forward to when the devs can integrate this model into the default Perplexity search available to everyone

[languid bronze]

Curious if perplexity has an API, came across https://platform.metaphor.systems/ who provide an API for llm based internet search and was curious if perplexity has any similar plans.

[barren scarab]

None so far.

[gray pumice]

Oh I was unaware. One has to je be registered tho. Right?

[dusk aspen]

Yeah, you have to be registered

[heady plover]

only requires an email, no credit card or anything

[dim bolt]

Would be awesome if developers could access Perplexity via an API

[silk aspen]

I just figured this out myself. Is it known why the extension is broken?

[dim apex]

is anyone here

perplexity is not working

[visual bough]

Hey, is it still not working for you? Can you please give more details?

[supple wraith]

perplexity also not working well in my case too! did I reach GPT4 rate limit? is there even such thing with perplexity?

[visual bough]

Hello @supple wraith, are you a Pro subscriber?
How exactly is there a restriction?

[supple wraith]

yes I'm. It's just stop responding from time to time

[visual bough]

The answer breaks off? Please share a few threads with this problem!

[soft silo]

congrats on new change of silently hiding the AI Profile section within Settings page 👏🏻

[unkempt tusk]

Not sure where to post this, but for some reason hitting enter on iPad to submit a query does not work — the text entry goes blank and I have to touch back in it and click the green arrow to submit.

[leaden hill]

what is the max lenght of the answer when in writing mode and analyzing 3-4 files uploaded, because very often it stops after two short paragraphs.

[pure canopy]

hi team, I've been wondering if this issue has been addressed directly? Prior to the finetuned model release I used perplexity many times per day (even if much slower). I had my qualms already with the quality and frequency of follow-ups, but the issues described here w/ the 3.5 release slowly brought me to the point of almost not using perplexity on a given week outside of a basic search here or there. the speed wasn't worth the degradation!

For example several times per day, copilot would ask me if I wanted to search for exactly what I had just entered, yes/no. had to prompt engineer just to avoid this, and for many queries realized I could just jump into chatgpt to ask a gpt-4 model directly to save lots of time (and that's what I've been doing this month. I miss pair coding with perplexity!)

I felt the 3.5 model could do waaay better with the right chaining! curious if you guys have been working on that

[visual bough]

for this case I actually don't know exactly, I guess 2k, but due to the file upload the limit could be reached earlier, maybe @covert sun can say more about this.

[pulsar ravine]

https://www.perplexity.ai/search/82a3f419-7599-4f73-ba3a-3897a2f431e9?s=m

Zagori isn't included in the lists even after it admits that it's a bottled water brand. Why does this happen

[dim apex]

it goes through an infinite loop, i cannot search anything from perplexity chrome extenison

[keen relic]

Hi, have use perplexity AI to search for software solutions, just tried one without checking virustotal , I become suspicious when the software didn't even support file format it was promoting to support, asked some detailed personal survey questions on install which was weird, and site looked completely legit offering professional ocr scanning services, but to my surprise the site had already been flagged by 7 security vendors, yet Perplexity Ai offered it up as the best option to my query for ocr scanner for image pdf files. Are there efforts in project to pre-filter / validate app recommendations by pinging some generally accept blacklist database?

[visual bough]

You tried this:

1. go to https://labs.perplexity.ai
2. solve the captcha
3. go back to https://perplexity.ai/

Not fixed?
Please disable the extension, the Chrome extension will receive an update on this problem in the future.

[dim apex]

i have tried this, it doesn't work. i have disabled and enabled the extenison. i have also unisntalled and reinstalled the extension no resolution.

[dim apex]

out of curioisty what llm does the extension run on?

[visual bough]

Can you use the perplexity when the extension is deactivated?

[dim apex]

https://www.perplexity.ai/ website works, when yo udisable the extension obviousl it won't work, i can't access the ext if extension is disabled?

[dim apex]

oh by hte way i am using microsoft edge

[visual bough]

There are known issues with the extension, a fix is being worked on, until then it is recommended to disable the extension!

[visual bough]

Pro users can change the model in their settings, for the other users it is the normal model!

[dim apex]

how would we know when to use it? i will forget to enable it

[visual bough]

It will certainly be announced!

[dim apex]

I would greatly appreciate it if you could always keep the legacy extension available to prevent any potential usability issues in the future. Having both the old and updated extensions at our disposal will ensure that we have a working option moving forward. This proactive approach will be highly beneficial.

[urban vale]

I think you should make a more affordable pro plan (5-12 dollars per month). I know gpt-4 is currently anything but cheap, but there could be a lower number of copilot uses per day (like 25-50)

[pastel ruin]

Thank God they finally updated the mobile app

[rocky shore]

Please add pdf support for the mobile app

[pastel ruin]

It's better but I'd still like the ability to see collections

[dim apex]

what was wrong with it?

[proper glacier]

I think everyone is waiting impatiently for that (including me)

[pastel ruin]

it was bad

you couldn't access library

all the posts were mushed together

the top was recommended searches instead of your history

[heady plover]

Hello, I need a reminder to buy the Pro package. But I want to ask if for the same question (not using Copilot), if perplexity enables GPT4 mode, will its answer be identical to ChatGpt Plus?

I wonder if GPT4 is the latest version from OpenAI, does it have any limitations compared to ChatGPT Plus in terms of logic?

[visual bough]

Perplexity uses GPT-4 via OpenAI's API, but I can't promise you exactly the same result (answers always vary).
A newer GPT model has not been published yet!

[coarse edge]

Anyone getting stuck on a page saying "Checking if the site connection is secure" on any URL on perplexity.ai ?

[glacial granite]

If you're stuck on Cloudflare check, please try this:

• go to https://labs.pplx.ai
• solve the captcha
• go back to https://perplexity.ai/

[coarse edge]

<@&1082051235542216704> Nope, no CAPTCHA to solve, just a spinning circle.

Here I had just opened the link for the Android app in the announcements forum.

Obviously I checked other sites, I have no problems connecting to those.

Tried closing and restarting my browser, all the hoopla, no luck.

[coarse edge]

After the 2nd try of the runaround you gave, again no CAPTCHA to solve, but things started working again. Thanks!

[long glen]

Hello team, having accumulated thousands of threads in my library I really miss a possibility to have an advanced search, not only through threads loaded on the page, but the whole history. For now, I found a solution with a small script that will automatically scroll through the infinite scroll till it finds a keyword, but it can take a long time...
Understandably, it's not an easy task for you to roll out advanced search for everybody, but if you had an API from where I cloud pull regularly all my library to an external DB with an index and being able to search with a 3rd party app it would be fantastic.

[glacial granite]

Hey @long glen! Perplexity AI does not currently provide an API.

[next acorn]

Can not access perplexity due to CloudFlare. Just keeps checking if my connection is secure.

Any ideas how to fix that?

[glacial granite]

If you're stuck on Cloudflare check, please try this:

• go to https://labs.pplx.ai
• solve the captcha
• go back to https://perplexity.ai/

[crimson nymph]

@next acorn ^

[visual bough]

@heady plover
You tried this, but it's still not working:

1. go to https://labs.perplexity.ai
2. solve the captcha
3. go back to https://perplexity.ai/ ?

[crimson nymph]

could you try disabling the perplexity extension and retry

the errors remaining are due to adblockers, no need to worry about that

the perplexity extension will receive an update soon that will hopefully fix those problems

[oak mauve]

Amazing creation, team! Long live Perplexity!

(finally I can stop using the super-bad Bing AI interface )

[abstract girder]

Yes this is how I've been solving it - but really? Every time? I'm going to have to give up on Perplexity sadly. First rule of UX - friction is bad!

[heady plover]

it's probably to avoid people from scraping perplexity answers

the android app doesn't have the issue but sadly no way to avoid it on PC

[crimson nymph]

Hi @abstract girder this is because of an issue with the perplexity browser extension. If you don’t want to follow those steps each time you could disable the extension for now.

[abstract girder]

Ah, so is this going to be fixed? I prompted a query in Perplexity about this and it said that the browser extension was dropped. I wasn't sure if this was a real result or not

[crimson nymph]

Yes it will be fixed next update that will be coming soon #🌐│feedback-web message

[abstract girder]

Great news! Thanks.

[next acorn]

This worked for me thanks a lot!

[delicate bay]

when would copilot credits i.e 600/day. but it is not coming back to 600/day, it is showing me 250/day even after 24 hours.

[visual bough]

The uses regenerate dynamically, that is, if you ask a question at 10 o'clock, you get this use back the next day at 10 o'clock.
So there is no fixed time at which you have 600 uses again.

[delicate bay]

@visual bough Thanks

[wet prairie]

Deleting threads is buggy, some of them pop up again. Also, it would be nice to display the delete button on the front (saves me 1 click per thread, which adds up time quite fast if you need to delete a lot of threads) 🙏

[karmic coyote]

Can't bypass the cloudflare sceen, it is constantly loading and not going forward. How to fix?

[barren scarab]

Oh mate, the mobile app is made for you. Scares the hell out of me because you can accidentally swipe and delete threads. But it sounds perfect for your use case. No confirmation needed, just swipe.
But mobile app loads only last 15 threads I believe.

[barren scarab]

#👉│feedback-general message

[karmic coyote]

thanks

guys, am I sure on Pro plan?

@covert sun

can't use AI models and seeing that try pro stuff

[covert sun]

Hi @karmic coyote! Did you try to reload the profile? Also, if you proceed to "Learn More" what do you see?

[karmic coyote]

I see the pricing tiers to upgrade, meaning I’m not on pro

There are only 5 copilots, I’m on free tier

Can you check my acc please? My email in the screen above

[cunning goblet]

@karmic coyote have you already subscribed to Pro, and still seeing this?

[jovial temple]

can't get answers to work on my ios device

says connection error, i've tried re-login, with and without vpn

works fine in the the web browser

[karmic coyote]

yep, I was on 7 days Trial Pro, like 3 days ago

Got the issue. All good, folks. My bad

[thin basalt]

very good point. perplexity needs to figure out how to counter nefarious people that know its going to scour the web looking for their articles

off the top of my head two solutions to design the ai to recognize data manipulation:

1. assigning far greater weight to results from "trusted/generic" sites (measured by things like alexa rank)
2. cross checking smaller websites vs their content to see if they have a bias

[delicate maple]

Hey! So me and a friend have been using the free perplexity and its great but we were using our five daily copilots and i used my five and sent the thread to him and he used his five and sent his thread to me and i got it back but only at the fifth message not the updated 10th that my friend had, any help?

[vapid mason]

bug: cannot type the / character in new thread box (causes the input to go to the greyed-out form below)

[crimson nymph]

they have this on their to-do list #1155837120129277963 message 👍

[cunning goblet]

Thanks will take a look

[vital arrow]

hello, can you add gift cards as payment option?

[oak mauve]

(Which also means I can also finally stop using the background-spying Edge browser on Windows entirely, forgot to mention that.)

[visual bough]

At Perplexity you can subscribe to Pro, subscriptions usually can't be paid with a gift card.

[valid pine]

Feature request: Threads that are not explicitly added into a collection are auto-deleted after X days.

[silk aspen]

Feature request: Threads give a suggestion to be added to a related collection, otherwise put into an archive queue to be auto deleted after X time. Similar to how unpinned tabs work in Arc browser

[crimson nymph]

This server is not for job search. If you are looking for a job. Perplexity has a number of open positions. You can view the job openings here. https://www.perplexity.ai/about

[crimson nymph]

@rough yew ☝️ ...

[manic shell]

My searches aren't finishing anymore. They just stall out. Any ideas?

[jovial temple]

check if websockets are blocked

if not try reconnect to internet and turn off any vpns if you have one

could try clear cookies, or a different browser to re-do the cloudflare captcha

[gaunt haven]

En Claude-2 cuando le pido que resuma o que me dé una parte del documento (que esta en inglés) me contesta en Inglés y yo tengo el AI Profile en Español.

[visual bough]

#👉│feedback-general message

[tranquil gate]

Hey, did it happen once or the issue is recurring? Thanks

[stuck iris]

My sidebar seems to be completely missing when using Perplexity on windows in Brave. Am I missing a shortcut or something?

[stuck iris]

Page is loading with sidebar display hidden

[manic shell]

Thanks, think it ended up being the chrome extension. I removed it and it started working again.

[visual bough]

Hey, can you please try again with all extensions and content blockers disabled?

[oak mauve]

Font size in the Android app is awkwardly small currently (on ~5" 1080x1920 displays), please add a setting for font size asap.

[quasi cave]

my problem with perplexity.ai is that: a) it lacks the re-generate button b) it seems to be overoptimized for generating summarization/short article c) it has trouble of generating lengthy article: it predicts eos token prematurely, leads to the sudden stop of generation d) if prompt contains multiple questions, it struggles to generate comprehensive answers in one go (compared to ask the question 1-by-1 e) sometimes the generated answer is empty f) it lacks the ability of letting user controls whether they want accuracy or diversity (sampling vs most probable)

[pulsar ravine]

https://www.perplexity.ai/search/30a858eb-b663-440e-bb91-750b77ba277c?s=u

Perplexity made a mistake. "The final episode, episode 30 of the entire series, aired on November 4, 2023, in Japan" 🤣

[barren scarab]

Don't think the thread is shared and you can flag the answer as well for being incorrect

[onyx swift]

#💬│general message

normal or bug after uploading a file and asking questions, after the second question perplexity switch the source to the web and completely ignore the uploaded the file for the rest of the conversation

[covert sun]

Hi @onyx swift! Could you share this thread, so we could take a look (you can DM it).

[hardy aspen]

Fake reference(s) in an answer from Perplexity AI. VERY unreliable and disappointing !

E.g.:
How come this statement "Optical systems: In optics, the propagation of light through a series of lenses and mirrors can be represented as matrix multiplication" is relating to this wrong reference?
https://www.scribbr.com/commonly-confused-words/anymore-vs-any-more/

Screenshot attached.

[onyx swift]

ok, I will send you the link

[alpine coyote]

A friend is trying perplexity but when asking any question it's not showing any answers, instead loading forever. Is anyone else experiencing this?

[trail iris]

Yes, the same problem.

[covert sun]

This was fixed, @onyx swift! Could you, please, check.

[flat anvil]

"Suggested Queries" thats trash.... same for Google I don't need "ideas", if I don't know what I have to do in the internet, then I leave and read a book or whatever...

#📢│announcements

[stuck iris]

With these new collection changes, can the AI see the description? Or will I need to repeat myself in the AI Prompt section?

[steel magnet]

Hey @stuck iris, good question! Right now we only use the AI Prompt because we wanted to make it clear when instructions are being used. We can look into adding description in the future

[covert sun]

Hey @alpine coyote! Is your friend seeing this on the web version or one of the apps?

[alpine coyote]

Hey @covert sun. On the web version. It seemed specifically tied to his account, because he tried on a private tab and it worked fine while he was unlogged

[covert sun]

Is your friend using any VPN services? DM me the IP if he does

[alpine coyote]

@covert sun Not as far as I know, but now I'm no longer present with him. I'll ask him if he faces the issue again

[stuck iris]

I've also had issues using perplexity on vpn

[cloud scarab]

Is it possible to ask Perplexity about how they came to a conclusion? To be more specific, when it quotes a specific source I want ask more about the specific source, which part of the wiki did you get that information from? Can you quote the source directly?

When I ask about the specific source it creates a new search with new sources. I think it would be useful to limit Perplexity to source 7 OR provide a specific source to use if that is possible.

[hazy mortar]

In the android app, the language of the responses is based on the settings in the computer, but those settings don't exist in the app.

Can they be added?

[covert sun]

Hey, @stuck iris! If it doesn't work on VPN, then you could try another hotspot/VPN service. Using it without VPN shouldn't have any issues.

[oak mauve]

Please allow us to remove the "8 most recent threads/collections you've visited" sidebar-list with some button click.

[oak mauve]

Since Firefox is among the most popular browsers, please also release the addon on https://addons.mozilla.org (for free!) and not only for Chromium-based browsers.

[barren scarab]

Would love this as well! Raised this about a month back but was informed that there are no plans as of now.

[fierce root]

It may help to reduce this error if you put to include credible references in their response. https://www.perplexity.ai/search/Explain-the-specific-ViW08p5VQmeBxYDF1KRCfA?s=c

[kindred arrow]

This thing just spews shit

It was great 2 months ago

whatever you did undo

it cant even tell me somthing as simple as windows file tree order

[kindred arrow]

https://www.perplexity.ai/search/Force-to-reset-maSG.BbJSIKES8cmSkQMSQ?s=c

My bot is literally trained better please switch the ting back i dont know why you guys did what you did

"I'm sorry, but I couldn't find any relevant search results for your question. It's possible that the query is too specific or that there is no direct answer to it. However, I can suggest some general steps that you can take to reset user permissions using the command prompt in Windows:" =major spew of worthless information all filler

so i have to read a paragraph of spew to then read more spew and then to read the wrong info

[visual bough]

Hey @kindred arrow!
You use the "Wolfram Alpha" focus, this is only useful for some special cases and otherwise causes hallucinations.
https://www.perplexity.ai/search/Force-to-reset-v7MtbqMjT8C6gpxSkgjAwA?s=c

[kindred arrow]

used all\

[kindred arrow]

yeah so what\

What I'm saying is I've used them all

It wouldn't matter what I used the thing just spews non-stop like it doesn't matter if he's alpha or not it still spews

[kindred arrow]

This is the only thing I can get out of it anymore whether I use any mode you can name it

[visual bough]

You used a focus that explicitly asks only Wolfram Alpha. And since Wolfram Alpha is not familiar with Windows, there would be a lot of halucination on Wolfram Alpha's part.
To prevent such a thing, you should not use a focus in this case.
But it could also be a bug, are you using Perplexity in the app or on the web?

[kindred arrow]

dude ive gotten worse results on easier stuff

web

[heady plover]

I wonder if AI Profile can be used to make the response shorter

and more to the point

[harsh rain]

Show the results with the other focuses

[quasi cave]

g) the inference speed is too slow, it wastes me alot of time to wait until the generation is completed

[heady plover]

I mean it's the fastest compared to competitors

but it can certainly be made faster by using open source models

[quasi cave]

summarization task is relative easy to perform, since it only has to condense the information, my use case is the inverse direction of summarization, e.g., makes a (incomplete) script/summarization comprehensive (with the help of web search)

[quasi cave]

about that, sometimes i got an error during the inference, it is really frustrating

[quasi cave]

here is my profile for claude 2:

We are academic experts.
We specialize in the following fields: data analysis, statistical learning theory, Bayesian learning, machine learning, computer vision, computer graphics, robotics, control theory, natural language processing, reinforcement learning, machine translation, and deep learning theory.
We use markdown and latex.
We answer in bullet point style with examples.
We answer with fine details.
We answer with contexts.
We answer in-depth.
We answer on all topics.
We give condensed and detailed answers.
We answer with examples.
We answer with explanations.
We answer with justifications.
We answer with theory.
We answer with practice considerations.
We answer with comprehensive text!

with profile: https://www.perplexity.ai/search/what-is-monte-kohx8EaGT1.bd_ZSfArPRA?s=c
https://www.perplexity.ai/search/what-is-monte-ndxQrUsWSvW3B1oPThir6A?s=c

without profile: https://www.perplexity.ai/search/what-is-monte-ukD6nLQBQbOxq6Nsd58syw?s=c
https://www.perplexity.ai/search/what-is-monte-o_PFg3YmQw..xbwht_4Ojg?s=c

this profile produces lengthy answers, to have a short response i added "given a condensed conclusion" in the prompt
with profile: https://www.perplexity.ai/search/what-is-monte-Rmy3xVFwRamPQHrn3YgdPA?s=c
https://www.perplexity.ai/search/what-is-monte-KWMipIaWQiG_OZBbIDhjsw?s=c

without profile: https://www.perplexity.ai/search/what-is-monte-5Blk4TlDT0mpac9p8cVfcg?s=c
https://www.perplexity.ai/search/what-is-monte-3wHS8M74R_WwCKP.Ru6cdw?s=c

[kindred arrow]

Switchy is an extension that can help boost conversion rates by using deep links strategically. Deep links can increase the likelihood that users will take desired actions within an app, such as making purchases or signing up for a service
1
. By leveraging psychology and optimizing landing pages, businesses can also increase their conversion rates
4
. However, it is important to track actual results and costs to ensure that campaigns are effective
2
3
. Additionally, businesses can improve their conversion rates by focusing on longer-form content and improving their website's traffic
5
6
.
why special
Sources

Answer
The term 😘"special".😒 is used in some of the search results in reference to deep links, but it is not clear why they are considered special. Here are some examples:
An Android App Link is a special type of deep link that allows your website URLs to immediately open the corresponding content in your Android app
1
.
Deep linking is just defining a route that takes you to a specific page, nothing complicated or special that you need to do when it comes to implementation
4
.
It gives you a special link to share. When someone clicks that link, it will open in the YouTube app
6
.
Based on these examples, it seems that "special" is used to describe certain types of deep links that have specific functionality or requirements. However, without more context it is difficult to determine exactly what makes these deep links special.```

user: why do i want it
Sources

Answer
```It is not clear from the search results what`"it"`

refers to in the question. However, based on the search results, here are some reasons why someone might want to use deep links or extensions:```

I am🫠 done i am embarrassed for ever telling people this was the best

i am embarrassed for the 6 other devs they didnt do this.

Its like your Challenged distant cousin with Asperger's ":Well certainly and well certainly " " If by bad you mean bad ass well then count me in"

Artificial Asperger's

[eager trail]

What day is it today

[hardy briar]

what is the difference between 'ask' and 'writing'?

[visual bough]

Do you mean the all and writing focus on the website?

[spare trail]

Do we report feature requests/changes as bugs, or is there a separate form for that?

[visual bough]

You can suggest new features in #💭│ideas, you can share your feedback for the web version in #🌐│feedback-web, for iOS in #🍏│feedback-ios, for android in #📱︱feedback-mobile and general feedback here.

[spare trail]

okay thanks!

[vital arrow]

hello, can you add search in specific collection?

[keen relic]

I'm still so excited about how Perplexity AI has transformed my life, work and personal. Hours and hours of time being saved, it is an edge I truly appreciate. 🥰 I share it with only those i care about, so once again, I send out my thanks to the Perplexity AI team.

Not just a time saver, it's a life tranformation tool, a way to change how we do things.

Just notice this:

Keep going, keep going., PAI team, thank you.

[hardy briar]

I mean the #ask channel and #writing channel in Discord

what is the difference?

[visual bough]

In #🔎│ask you can ask anything, in #✍│writing you can ask Perplexity and/or other users about writing tasks.

[hardy briar]

thx

[harsh rain]

I understand now

Recently they have been adding references

To replies I think

I don't think the fine tuning was there before

https://www.perplexity.ai/search/Sample-github-axtions-p7jJMjmYSSGnvvSSeDlnfA?s=c

Example replicating it

Maybe it ran out of context?

[pure leaf]

Where is the best place to request tech support for a payment issue?

[visual bough]

What exactly is the problem?

[pure leaf]

I tried subscribing to the annual plan but changed my mind. It failed to bill the annual amount and now I can't use that card to subscribe to the monthly plan because there was failed payment on it. I don't understand because I can easily transfer money but that is blocked in my account.

[visual bough]

In this case the best is to contact the support at [e-mail]

[pure leaf]

I will thank you for your help!

[sweet palm]

if i upload 3 documents in a thread in a collection, it sees only the first files and takes info only on that.

[tight dragon]

I'm also experimenting with pdf files. Yesterday I uploaded two files and only the last file was read, remembered or contextualized. I believe in my case that was due to the length.... By the time it starts the second file, it has already exceeded the context length, so it cannot remember the previous file. In my case that makes sense because they're large files. In your case how large are the files?

[sweet palm]

25/35 pages NIST regulations ( NIST.IR.8259, NIST.IR.8259A, NIST.IR.8259B)

[tight dragon]

That's not too long. I don't know man I'm also new, but I'd like to know what the team says about. I also uploaded more than one file but only one was read ...

[sweet palm]

yeah i also asked "how many file do you see?" and perplexity answered that it could see 1 file (and then said the title of only one of the three)

[tight dragon]

You could merge the three PDF's in one file and upload it...

[tight dragon]

#📘│faq message

[tight dragon]

#📱︱feedback-mobile I imagine that the Android app is still under development, but these are the things that I noticed:

• after creating a thread in the web version the list is not updated in the app
• there is no tab for collections, I cannot upload files and there is no AI profile in settings
  Other than that it is very good. 🍻
  I'm using Android 13 and Chrome. 🤖

[pearl wraith]

Error occurs when using llama replit option

[covert sun]

Thanks for reporting, @pearl wraith! We'll investigate

[rich barn]

Can I ask if I used a promo code when subscribing to Pro, and if I renew it later, will it revert to the original price or continue at the first discount?

[crimson nymph]

the promo code is 1 time use only

[rich barn]

Okay. 😟

[keen relic]

When I have a thread going on 'how to' procedures/steps, I've noticed that when posting followup questions to expand on previous response, the first part of response is to provide a paraphrase about the previous response in the current response, I understand that it is continuing the discussion, but the amount of re-hash is too much imo, it could be reduced, or a reference statement made, like "in continuance of our previous discussion above"

[tight dragon]

It is awesome that you can ask perplexity not to do a web search and it won't. One doesn't need to specify a writing scope necessarily. Cheers for that ! 👍🏼

[shy forge]

Premise : some websites need to login to surf through more information.
New feature suggestion : Integrate with websites like Reddit/twitter/LinkedIn/blind/etc…. Have a new UI to the user where they can input user name and password for each of these websites beforehand.
These credentials can be used by perplexity to login onto the specific websites and surf through them for more information.

[analog shard]

Why Claud AI Can't Work Is it offline

[mortal echo]

in general, when I write a request and press "/", the printing starts elsewhere, the cursor is not in the active field

[visual bough]

https://discordapp.com/channels/1047197230748151888/1155837120129277963

[mortal echo]

has it not been fixed yet ?

[visual bough]

No, a fix is being worked on!

[radiant kindle]

https://www.perplexity.ai/search/a6387681-4b50-474b-a24c-ee74e3f73d74?s=m

I was asking whether "mold" has another meaning

it answered with meaning of "clarification" instead

I seriously wonder what was in the AI's context memory

[quasi cave]

can perplexity.ai provide the confidence level of the output in the output? this might be helpful for not misleading the user with wrong information (assuming the confidence level reflexes more or less the truth)

[green bear]

This is a bit confusing. I wanted to check whether it would be possible to use GPT4 for a writing task, but the model tells me it is based on GPT3. I am unsure whether this is actually the case, or that the system prompt is simply incorrect.

[visual bough]

It is GPT-4, but GPT-4 does not know that it is GPT-4 because at the time of training there was no GPT-4 yet.

[green bear]

Oh okay gotcha

As some general feedback I would like to add.

1. Amazing tool, I tell everyone about Perplexity and share answers using the URL to many people. It saves so much time. Props to the team behind this tool!
2. It would be great if it would be possible to switch to writing mode in an existing thread to help rewrite the text in a different format.

The initial questions that I use are often investigative, but the answers are not ready to be shared. After asking for example three questions, it could be useful to be able to ask Perplexity to write a summary of the answers so far, or to write a short article about it. As far as I know it might execute an unwanted search when doing this.

[tight dragon]

Hello, I don't know if this is the right section, but when I asked a question relative to one search results, Copilote asked me back the same question:
https://www.perplexity.ai/search/I-think-metal-cDJ7w7dvQQW5wB4fw_7KxA?s=c

My question wasn't very well formulated maybe.

[green bear]

I experience this occasionally as well. The question that co-pilot asks tends to be open and formulated almost the same as the input.

[opal isle]

Hello, is it bug ? i cant copy part of the code generated 😅

[sour acorn]

#1162701258826715297

[karmic edge]

Whenever I mention "weather" in a query I get a forecast for any city that matches part of the prompt

[coarse edge]

Hello, I found a problem with the Android app:

When I ask a follow-up question spanning multiple lines, eventually the 'send' button disappears, and I am unable to send my question to Perplexity.

(I use large fonts because of my vision impairment. This may be part of the problem.)

In these two screenshots you can see that the 'send' button disappears as soon as there are several lines of text in the text box and I manually start a new line.
The same problem occurs when the text I enter wraps and takes up multiple lines.

Steps to reproduce:

• in the Android app, start typing a followup question. keep adding lines and check if the 'send' button disappears.

Additional observations:

• When the 'send' button has disappeared, hitting the back button in an attempt to hide the keyboard or switching to another app and back to Perplexity discards any text that was entered.

[coarse edge]

The problem appears to be even worse: when I type anything in the "Ask follow-up ..." text box and then before I hit 'send' I switch to another app and back to Perplexity, all entered text disappears. This is very inconvenient when I want to copy some text from another app to send to Perplexity.

[hollow sail]

Is there a reason you can't copy individual code lines anymore?

Seems to have been broken for ~3 days for me

[visual bough]

https://discord.com/channels/1047197230748151888/1162701258826715297

[oak mauve]

Please post a statement regarding which GPT model Perplexity uses. As GPT-4 is selectable in this menu (and labeled with "pro"), it appears that Perplexity does not use GPT-4 by default.

[visual bough]

You can select the model in the settings (Perplexity, GPT-4, Claude 2).
The feature you are referring to is the rewrite function with which you can create the answer again.
You can also see which model is used at the top left of the answer.

[oak mauve]

When entering quite long thread names, they are shortened by "..." and do not fully appear when you hover the mouse over the thread name. Please make the full thread name appear in the hover-context-menu when hovering the mouse over the thread name.

[oak mauve]

You can also see which model is used at the top left of the answer.
no I don't see that 🤔

[visual bough]

This indicates you are using Perplexity model

[visual bough]

Have you even set another model in the settings?

[oak mauve]

I clicked on Copilot in that "rewrite" menu, yes.

[visual bough]

No, I meant in the settings 🙂

[oak mauve]

Oooooh! Found it now, sorry. No, I have not set anything else. Also, that setting is greyed out for me, stuck at "Perplexity" setting.

[visual bough]

This option is only for Pro subscribers.

[oak mauve]

OOooooh, one free GPT-4 request every 4 hours, aaaah, interesting, thanks!!! Now it's clear!

👌

amazing features! ( is my own custom discord server emoji 😁 I created it just now)

[visual bough]

Yeah you get 5 queries/4h.

[oak mauve]

okay, interesting concept!

[oak mauve]

Please add a timer so users can see when the 4 hours are going to be over

[brave frigate]

So it would even easier to use without paying for pro.

[fierce root]

Whenever I click on the drop-down menu to rewrite my answer to the initial query and select one of the options given (Copilot, GPT-4, or Claude-2), I noticed that no copy of the original response generated was saved. To compare the results of all the models, it would be helpful if there was an archive or record of the different replies provided by each model in the thread.

[lofty dawn]

Recently, there has been a decline in performance during the same timeframe. As a result, should Perplexity consider expanding its capacity to handle more requests?

[pearl wraith]

There are issues with the Wolfram focus. Additionally, perplexity seems to copy and paste any word problem or randomly paste its response to the user into Wolfram, instead of assembling appropriate math prompts for follow-up. It needs to be very similar to how GPT-4 handles users' math queries with the Wolfram plugin.

[trail verge]

I have been with Perplexity for a while now, it absolutely feels like the quality of responses has been and is taking a serious dive lately. I used to have great conversations that made me want to ask more questions, lately I am just getting a summary of the google summary, like literally the most low effort responses that I would expect from asking a middle school student. I used to get technical details, that has gone away entirely. Heck it used to try to answer, now it just recommends I go to a hired service. Context has gone completely out the window as well. Follow up questions seem to have almost nothing to do with previous questions and I find myself constantly repeating what it is that we're talking about. Pretty much my experience with AI has gone from a high proficiency assistant with vast knowledge available to a middle schooler with ADHD and google.

I am not sure if Pro is worth it anymore, I loved it when I first got it, but am finding myself more and more just conducting my own search cause perplexity won't stop mirroring all the AI generated tech articles. It just feels like an echo chamber for low effort content.

[tight dragon]

Could you share some examples?

[tight dragon]

Personally, I'm 300%-400% more efficient in my work. I just finished in 1 hour what used to take me 4. I suppose if you want to get the best or if this tech you also have to adapt the way you do your work. May I ask what is it that you do?

[unborn storm]

Hi, I've got a suggestion for the app version of Perplexity, specifically I'm using Android. On the Desktop version, I use the 'collections' and 'threads' feature a lot, for example to group different subjects together, and find it super helpful to organize my work. On the app however I can only really see my most recent threads, so it gets a bit messy and confusing from time to time. Is it possible to implement Collections into the app as well?

In addition, I personally think it would be a nice touch if all the threads that belong to one collections were sort of indented below the title of the collection (like a folder substructure/heirarchy), just to organize up a bit more. But this is more of just a pedantic opinion of mine!

[tight dragon]

Like a WYSIWYG layout editor ? I like it how it is, but if I'm searching something I'd prefer it to be right below the title as you say.

[tight dragon]

My understanding is that they are currently working on that, we must be patient. I'm using the web version in Android in the meanwhile.

[oak mauve]

Please integrate LLaMA 2 (from https://labs.perplexity.ai ) into the Perplexity Android app. I'd love to use a more energy-efficient language model than GPT-3.5 with nearly as good results as GPT 3.5 (which LLaMA 2 apparently is. Let's do sth. good for the environment and allow us to save some energy!)

[oak mauve]

Please make labs.perplexity.ai save at least the last submitted query+reply when closing the browser and not deleting the browser-cookies upon closing.

[marble monolith]

does perplexity ai use chatgpt when out of the copilot mode?

[tight dragon]

I dont think so, there's no API for ChayGPT to my knowledge, only for GPT 3+

[tight dragon]

Also the use of model is a separate setting, Copilote only asks for clarifications independently of the model.

[tight dragon]

So yes, if you have gpt-4 in profile settings (other than Perplexity ai and Claude 2).

[sudden dove]

Request to stop the spam pop up to install app (I already did this) every time starting a new question

I talk about the browser version

[sudden dove]

I noticed perplexity became less intelligent today, or is it my imagination?

[vestal sphinx]

There is though, ChatGPT is available as gpt-3.5-turbo on the OpenAI api

[lofty dawn]

In the past week, we have experienced connectivity issues with your product in the Hong Kong region, with no response between the hours of 8 pm to 8 am. Please consider enhancing server throughput and optimizing the provided premium services.

[tight dragon]

It's confusing, I though chatGPT was another thing than GPT3. Thanks for clarifying.

[upbeat sinew]

No, I noticed that too! It is pretty much useless now 😦 Hope they fix that soon!

[tight dragon]

In the web desktop version, if you paste a long text, it converts it to a paste.txt file 😐
That's amazing ! 😀
I'm using Android mostly so I miss those cool features, but the app it's already good as it is. But don't hesitate to add those features to the app 👍🏼

[vestal sphinx]

You're welcome!

[sudden dove]

What happened? Does anyone know? Seems to get very easily confused now

[upbeat sinew]

Nobody seems to know unfortunately 😦 I submitted a bug report #1165273167334494318 message

[sudden dove]

I hope this AI did not have to be lobotomized for political reasons 😄

Sometimes it even answers me in a different language without prompting.

I thought not starting a new thread was the problem... Getting confused with previous inputs. But no

[upbeat sinew]

Same 😦

[rancid ginkgo]

Perplexity is borderline unusable now. wtf happened?

[ashen cape]

Why make the price of premium same as ChatGPT pro?

[lyric dock]

ChatGPT Plus is $240/year

[ashen cape]

i mean for the monthly plan 😅

[lyric dock]

I think it might be because of the cost of GPT-4 🤔

The price also includes Claude-2

[kindred orbit]

I noticed it, too. Setting the AI model from "perplexity" to "GPT-4" in the settings seems to make it better again. Could it be that copilot has been switched to being "perplexity" (GPT-3.5 FT) for both information gathering and the final answer? 🤔

[upbeat sinew]

Thanks, I noticed that too! It seems that, using "Perplexity" as model, llama is now being used (at least for my account) - I tried comparing the outputs with the ones from https://labs.perplexity.ai/ and it seems to confirm my hypothesis

[kindred orbit]

Interesting, I haven't been to the labs for some time and now this is new (see picture). Yet another hint towards my theory that perplexity model is becoming LLAMA 😆

[upbeat sinew]

This might be it then! It's just a pity that performance is significantly worse now 😦

[gentle mantle]

I think the co-pilot on iOS is broken because It always says to check the pilot count left, even those I tried to reinstall. I think this happened with my last updated

[visual bough]

It works for me, what version are you using? Do you use a VPN?

[kindred orbit]

I wasn't sure if it's really worse but after playing with it more today, I can say: yes, it's really worse. The answers are formatted worse, you mostly just get a "wall of text" and what's even worse: it messes up facts even though the answer is right in the source it links. The perplexity team should really be more transparent on this. Don't know how I feel about it. On one hand I really would like LLAMA to work for them (more control over fine tuning and infrastructure) but on the other hand currently, it just isn't the same quality as OpenAI's models. When I signed up for Pro (yearly subscription), copilot was all GPT-4, then it was changed to GPT-3.5 FT + GPT-4 (at least with announcement) and now seemingly completely moved to be all LLAMA with no communication and no really transparent way to compare and provide unbiased feedback. As a former Bing AI User, I signed up because I was really convinced by the accuracy - it was the first time I could rely on the answer and didn't have to question anything, but now it seems we're back again at not being able to trust the AI. 🙁

[upbeat sinew]

I agree, as much of a fan as I am of Perplexity and Copilot, I feel like I cannot trust it anymore, and also I feel like that this lack of transparency is unfair to us pro users 😦

[tight dragon]

In my opinion nothing is perfect, and sometimes there are periods of difficulties, I think the dev team must be aware of this, I wouldn't lose my trust just because of what in my opinion is a temporary flaw, especially since this is just the beginning of AI and it is to be expected some trials and errors.

First time in human history we're doing this, give them a break.

That being said I wanted to report another flaw, which is that when I ask it to correct a text, it is not correcting the text that comes afterwards but the text of a previous message.

(I can share the thread privately if someone wants to have a look)

[gaunt haven]

The only thing I find disappointing is that you can't use the entire contents of the files because Perplexity summarizes if the file is large. So when I ask about parts of the content it doesn't give me the correct answer.

[chrome nacelle]

I would like to report a problem. When I click "rewrite" with Claude-2 in Copilot mode, after initially searching with GPT-4, it is unable to query Internet sources. The search results are empty.

[sudden dove]

What happened to this AI, it is much worse suddenly. Started a few days ago

It is almost unusable now...

Unsuccessfully tried to force perplexity to do a simple task several times... Randomize the order of the questions, remove text in brackets, don't answer the questions, leave the blanks.

With chat gpt on bing it did it first time, no problem. I feel like perplexity could have done this task before recent changes a few days ago

[lofty dawn]

Today, I conducted an experiment that demonstrates the superior stability of web-based applications compared to mobile applications, particularly on Apple phones.

As a paying customer, I have now also transitioned to using the web-based application.

I hope that you will continue to improve the stability of the mobile application's connectivity.

[upbeat sinew]

I feel the same! If you have pro, switching to GPT-4 helps (even though it does not resolve this completely).
I opened a bug report here: https://discord.com/channels/1047197230748151888/1165273167334494318 I think that you could share some examples too, if you want, in order to help them better understand the problem 🙂

[robust cypress]

Hi. I need to change an email but I have paid for a yearly subscription. Can someone support?

[visual bough]

Changing the email via the interface is currently not supported, but you can contact support ([e-mail]) 🙂

[robust cypress]

I no longer have access to the email I set up my account with.
Can I contact it from a new email?

[visual bough]

Yes, but please mention the old email address

[sudden dove]

I should write an example in your bug report?

I don't have pro. Just the free web browser version I am using

[upbeat sinew]

If you want to, you may share them as a link as they requested in that thread 🙂

[slim hinge]

Wasnt doing this yesterday 😭

[final glade]

The AI Prompt in every collection didn’t work

[oak mauve]

Please store the last language-model selection that the user used on labs.perplexity.ai in a browser-cookie so that we don't have to change the language-model (drop-down list) everytime we open the page (those of us who use a model other than mistral, for example the open-source "llama". "mistral" is selected by default every time).

[blazing epoch]

This has been a pretty common issue more recently.

It's either that or the response it gives is completely blank lmao

[visual bough]

Hey @potent ermine!
Regarding the data loss, please contact [e-mail].
Regarding the inaccurate answers, it would be helpful if you could provide some examples (with links). You can also report inaccurate answers to the AI team at any time using the flag icon below the question.
Thanks in advance!

[oak mauve]

labs.perplexity.ai website is quite cluttered, readability of the reply area is quite bad, due to the big language model selection area being quite big and in the middle of the screen instead of on the bottom.

[raven nacelle]

I’m not getting a response to asking anything. It’s just sitting there. Is something being worked on?

[sand geyser]

perplexity as of right now seems to have lost all ability to read uploaded files.

[crimson nymph]

Does seem to be working. Make sure you have the focus mode on writing. If you still have this problem after trying that please create a thread in #1140622008086970420 and share your perplexity thread where this problem occurs, this way the team can better help you with your problem.

[sand geyser]

Here take a look. It is still failing. Anyway I will bring it up with bug reports.

On second thought I can’t bring it up with bug reports. Infosec issues. Oh well.

I switched to gpt-4 and I switched filetype from pdf to txt. Same problem.

[crimson nymph]

Could you turn off copilot and check if you still have this issue

[sand geyser]

Ha! It works! Thanks. I’ll stick to this till the problem gets fixed.

[hollow jay]

Im confused, I’m trying to use the app today but it claims I’m not logged in even though I am ??? And it claimed my username was invalid or something

[visual bough]

Please check if you are using the latest version, if there is no update, please reinstall the app.

[hollow jay]

Ok thanks

[south ether]

Can you DM me the URL link to this? you can make it public by clicking on the top right "Share" button and making it shareable, thanks!

[sand geyser]

It looks like copilot is working again with file reading. Thanks again for the suggestion to turn off copilot.

I'm still figuring out copilot. While I doubt it's much help in file reading, when it comes to web searches, a lot of the time it seems to make little difference, but there are a few times when it massively improves the result.

[ashen hinge]

can someone explain to me why every language model is outdated in terms of information?

from GPT-2 -3- and 4

to Llama

[quasi cave]

what went wrong? is there a timeout or max token length limit?

[visual bough]

These are usually just isolated glitches, if you ask your question again it should work.

[quasi cave]

i tried multiple (~10) times

i got this problem every single time

[visual bough]

Please also refresh the tab

[quasi cave]

i always ran the prompt in a new tab

[quasi cave]

i split the input pdf in half and the problem resolved, this highly suggests there is a hard cap on context length

[south grove]

Ok, so I have a request. If Perplexity returns a blog with 10 answers/sites - don't link to the blog, link to the direct sites.

Basically, there are a ton of SEO honeypot pages that HAS THEIR PRODUCT MENTIONED IN THE LIST. Making it garbage. And now that most of these are AI generated, it's even worse. I'd like Perplexity to ALWAYS give me the option of a direct link.

Posslbe?

[tight dragon]

Each model is trained with past information, so it will always have outdated information until the day we create an AI capable of seeing the future; then we'll have up-to-date information (half serious, half joking answer).

Ah, or until now that we have Perplexity so we can have updated information..

[ashen hinge]

Oh interesting but at least we have perplexity though

perplexity is like Google and ChatGPT combined..

Nothing like it..

[tight dragon]

Bing also makes web searches, but it's very oriented. Sometimes their ai refuses to continue and closes the conversation ...

[woeful sparrow]

Seriously @heady plover ? At this point this is blatant failure of your basic core feature..

Perplexity is now unable to perfom web searches..

[warped briar]

did this not used to be the case? I haven't used Perplexity much myself, yet, sounds a bit like a deal breaker

[woeful sparrow]

That is the basic product promise of perplexity, allowing you to search the web using an AI assistant

between this, the low quality answer, the copilot question that just don't make any sense, and the fact that Open AI Bing features now complelty outperforms perplexity

I used to be completly mindblown by perplexity abilities but it has seemed to go downhill

[visual bough]

Hey @woeful sparrow!
This is not always the case. Errors can occur, just like with any other product, but most of the time it works.
Please share a link to this thread so it can be investigated what the problem was and it can be fixed.
Thanks in advance!

[woeful sparrow]

Hey, sorry, I would rather not share the thread publicly for privacy/confidentiality reasons, however I could share it privately with Perplexity team if asked.

[visual bough]

Feel free to send a DM, for example to @covert sun

[frozen scaffold]

Just a simple feedback to say Perplexity is genius and I use it almost everyday for pro and personnal purpose 😎

[whole spear]

please provide dark mode on web perplexity thankss

[shadow moth]

今天安卓版本总是提示无法登录

[visual bough]

Hey @shadow moth, please provide more information about your problem, like the Android version and the app version, also a screen capture would be helpful.
Thanks in advance!

[radiant kindle]

It's so frustrating when the AI misunderstands my question
me: what is sentry
Perplexity: (finds a few things called "sentry")
me: what is the base, dictionary meaning of the word
I clearly meant meaning of "sentry", but perplexity: (meaning of "base")
(no link to the thread available, I edited the question)

[ashen hinge]

What are you talking about? Perplexity automatically searches the web and provides sources for every answer..

[woeful sparrow]

Well look at the screenshot above

It didn’t provide any link or source and just told me it couldn’t search the web

[ashen hinge]

That Is probably just an error try it again..

[ashen hinge]

Perplexity always provides sources

Stop acting like it doesn't

[lyric jacinth]

quick feedback vent:

please make the pro surveys less invasive by taking out of feed. I've taken like two surveys just recently already and it should resolve to the backend and be on cool down for a while or something.

android mobile app keyboard interaction has burned me a few times now, but it might be because I use swiftkey (ms alternative keyboard on android). I've lost a few impromptu prompt paragraphs that were decently refined from having it just vanish when I go to collapse the keyboard.

Other than that ❤️ perplexity. Fantastic work.

[heady plover]

Perplexity widget went blank on iOS 17.0.3. Clicking on it still opens the app. Still shows up as blank when I delete and create again

[oak mauve]

Please make Perplexity detect if a query even contains a question or task. It's fully unacceptable that it responds with a long text onto a dumb query like "Challenges in AI integration in supply chain: Bias and fairness in AI decision making" that lacks ANY question or TASK!!!!

(source: <#🔎│ask message>)

[oak mauve]

Sounds more like a post for #1140622008086970420 than #👉│feedback-general ...

[heady plover]

good call 🙏

[subtle flare]

What happened to my subscription? It did not renew, I was notified, account set to basic. Can not access any of the billing functions as the account is no longer pro. Can not “upgrade” as I already have an account? There is. “Restore” function in the mobile app, but doesn’t work. However it looks like I can pay through Apple Store.

[pulsar ravine]

Thank you, I made it shareable. Is deleting a thread after flagging it a problem? I doubt it would but I just wanted to make sure

[mild vector]

Perplexity does not appear as an option to choose as the Digital Assistant app on Android 14 (Pixel 8).

Is that a bug or just not something you've implemented?

[covert sun]

Hey @subtle flare! Regarding the billing issues, could you please send more details to [e-mail]. Thank you!

[subtle flare]

I immediately did that. There has been no reply.

[covert sun]

Thank you! We'll check.

[lyric nimbus]

hello, I got this strange question from copilot. for the context - czech has three grammatical genders (maskulinum, femininum, neutrum). is it asking me to assign "on" and "off" genders to words, or what is it thinking? 😄

[oak mauve]

Please add a "bright color mode" (dark letters on white background mode) for the labs.perplexity.ai website.

[visual bough]

Hey, @rough yew!
Wrong server, please refrain from posting such messages.

[azure fox]

I just moved from OpenAI to Perplexity AI. One thing I immediately notice is that the browser does not scroll down as the output is generated. Is this possible to implement?

[azure fox]

I think it’s about time to replace CodeLlama with Phind perhaps. It performs much better from the looks of it and I would definitely use it myself.

https://huggingface.co/Phind/Phind-CodeLlama-34B-v2

[potent silo]

Does Perplexity support the multimodal aspects of GPT4 yet? It'd be nice to also input voice or pictures... As well as ask for the output of audio and pictures... The GPT4 API supports this I believe?

[visual bough]

#💬│general message

[pure vigil]

Something I notice is that Perplexity ONLY uses web results to give information. It doesn’t utilize the ChatGPT knowledge. I realized this when asking Perplexity something, it didn’t have an answer for me. Then I went to the actual ChatGPT by open Ai (not connected to internet) and it gave me the answer I was looking for. Can you guys use both web results AND ChatGPT’s knowledge to answer questions? Thanks.

[lofty dawn]

Great observation! GPT-4 appears to act as a textual beautifier. However, when using “writing”, the content should be limited to the knowledge contained within GPT-4.

[barren scarab]

Issue seems similar to this bug report of mine. Could you highlight your experience and share a few threads for them to debug?
https://discord.com/channels/1047197230748151888/1157967736866091028

[ashen hinge]

hey im wondering what the Prompt option on a collection does? because it doesn't seem to have any power or take anything into affect when i type anything in..

[versed peak]

When will oled dark mode happen? The grey looks awful

[glossy heath]

DELETE ALL THREADS: This option is placed and warned a bit too lightly. I just accidentally deleted ALL my saved threads.
I was trying to just delete one, unthinkingly, I hit the wrong "..." and clicked yes (believing I was confirming the deletion of a single thread) and now they are ALL gone.

For a Delete All Threads, you should have to type DELETE ALL. That's my suggestions.

Is there any way I can recover these (another nice option)?

[ashen hinge]

exactly i dont know why they have that option

[glossy heath]

Well, I get why they have it. I think Claude is pretty ridiculous... You can't delete any chats... I just don't get why Perplexity implemented Delete All Chats with such little thought...

For ChatGPT you have to go to Settings > General > Clear all chats and click the huge RED button.

For Perplexity, you have "..." for Delete All Cats, just like you have "..." to Delete a Single Chat, and both are on the same screen, as easily accessible as one another.

Not intuitive.

[raw thunder]

I think delete all threads is super useful for my usecases.
I often clear them when I'm done with my research and save some of the highlights into separate notes, etc.

[ashen hinge]

why research certain things just to delete it all?

makes no sense

[random pine]

Yea true what I would do is that I would type down the names of every thread so I could then remember them or take a picture of every question and answer within every thread so you can then remember them. Good thing for perplexity it is different from like a notes or presentation for which you have to type in information when it comes to perplexity all you got to do is ask a question and it will give you the information so even if you loose it you can get the information back again by asking the same question, and it is not like you put that information into the threads and you need to type it all over again you know?

[glossy heath]

Kudos to Perplexity for restoring my deleted threads!

NOTE: Please make it more difficult to DELETE ALL THREADS 🙂 Thanks.

[raw thunder]

Thanks for pointing out.
I think perplexity is very good at reproducing its result whenever I want if I remember the query (or even similar query).
And that's the reason why I delete almost all threads after my daily job (which made me to research stuffs) - When the job is over most of the threads quickly become irrelevant to me.

[south ether]

You are welcome, just to be clear, did this happen for you on the app or web desktop?

[jade badge]

Hi,
I am currently experiencing difficulties with P bot (both web and Android) as it consistently provides poor sources. Even when explicitly requesting it to use a specific site, such as an online encyclopedia and not Wikipedia, it does not adhere to the instruction. Moreover, it frequently presents irrelevant sources that are unrelated to the topic. When prompted to reevaluate the sources, it often repeats the same errors. Occasionally, it includes a link in the response; however, these links are often dead.

Twice, I attempted to locate the articles by searching for the given link using the article's name and the site's name on Google, but to no avail. It is noteworthy that the sites where the articles are supposed to be located are within my country, so location should not be an issue.

Strangely, I feel that the quality of the answers provided by P bot was better during the first week of usage

[south ether]

Oh no, I'm sorry to hear the responses you are getting aren't providing the right sources. Is it often that it returns dead link sources? Unfortunately, it must be stated that Perplexity is not quite yet able to understand specific site requests in a consistent and reliable way. Are you using a regular search, have you tried with Copilot?

We want to take a closer look if you wouldn't mind sending a DM to me with a few threads where this is the case that would be extremely helpful.

[jade badge]

Thank you for your swift response. This is not an issue when using copilot. I have rapported a couple of the incidents. But I`ll DM you.

[primal violet]

Hi, android app is really slow. 3-4 seconds minimum starting time.

I rather go to browser perplexity and search fast there instead of using the app

Also, in the Web (Desktop), it has a small refresh time automatically. SO if I write anything in the first second, it automatically gets erased which is bit frustrating. See video. It happens everytime.

[teal geode]

Earlier this week I learned about "GPTs" and a tool called Mindstudio that helps you create them. Does Perplexity have a way to create these little apps?

[copper rune]

@covert sun while the “Add to Collection” feature is useful, I believe that it can be improved by implementing AI to make the process more efficient and user-friendly. Here are some suggestions on how AI can be used to enhance the "Add to Collection" feature:

• Automated Collection Creation: Instead of users having to manually create a collection and add items to it, AI can automatically create a collection based on the type of content being added. For example, if a user adds a news article, the AI can create a collection called "News Articles" and add the article to it.

• Automated Title and Description Generation: AI can analyze the content being added and generate a title and description for the collection. This will save users time and effort in coming up with a title and description themselves.

• Automated Icon Generation: AI can also generate an icon for the collection based on the content being added. For example, if a user adds a recipe, the AI can generate an icon of a cooking pot or utensils.

I believe that implementing these AI features will make the "Add to Collection" process more efficient and user-friendly. Additionally, I suggest renaming the feature to something that reflects the use of AI, such as "AI Collection Assistant" or "Smart Collection Manager".

[slim hinge]

Unfortunately, I have been experiencing multiple issues with my account, particularly when it comes to not being able to login and generate responses. I find this frustrating, as I rely on these capabilities for my professional and personal work. Not having access to these features has caused significant disruptions in my projects and creative endeavors.

I am aware that there was a recent update to your platform and I suspect that may have caused these issues. I would like to know how you plan to address and potentially resolve these issues in a timely manner. I am a long-time user of your platform and have grown accustomed to relying on its features for my work. Not being able to fully utilize these capabilities has caused significant setbacks in my workflow and I would like to see these issues resolved as soon as possible.

When do you expect a resolution to these issues and can you provide additional information about the cause of the problems? I appreciate your attention to this matter and I look forward to a prompt and satisfactory solution.

[queen flint]

I have a similar issue with my account too.

[slim hinge]

Yeah really annoying

[heady plover]

I have a Perplexity Pro subscription, and I am also having this issue. I can't even sign into my account, let alone use perplexity. I keep getting the "Try signing in with a different account." error when I try to sign in through google, and when I try to sign in through email, it says "Error." Someone please fix this or I will no longer be paying for Perplexity Pro if I can't even use it.

[slim hinge]

That is true and i hope theyll fix it

[woeful parrot]

I just joined Pro today and then it crashed while I was working. I hope they get the server back up soon, it's kind of frustrating.

[covert sun]

Should be back up now. Sorry for the inconvenience.
#🔔│status message

[frank aurora]

the GPT model of Perplexity Pro is GPT3 or GPT4? Why did it answer me using GPT3 when I asked it a question?

[heady plover]

It would be awesome if the question window would always show the current AI Model that is selected. Right now you either have to go to the settings and check which AI Model (GPT-4, Claude 2 or Perplexity) is currently selected or you have to wait to finish the answer to see which model was actually used. I find this quite inconvenient and a bit frustrating, especially when frequently changing the underlying model for different questions. Ideally you should be able to always see below the question window which AI Model is currently set and be able to change it right there.

[sour acorn]

https://discord.com/channels/1047197230748151888/1145894530239975474

[barren scarab]

Related request
https://discord.com/channels/1047197230748151888/1155593813998452838

[thin ginkgo]

Your web is amazing for my normal searches. Better than web browsing

[night meteor]

If you need to search for some trivial info its. If for something more serious - its unusable.

[crimson nymph]

Could you share an example thread this will help the team to improve the output. You can also report the the answer at the bottom of the thread by clicking on the flag icon if the answer isn't helpful.

[night meteor]

i already sent you feedback to bot.

[rugged flare]

what's going on with perplexity today?

it's slow, and hallucinating from time to time

totally not on the subject...

[pulsar ravine]

Please make bulk deletion a thing, or just make deleting threads less time consuming!
Edit: just like holding shift over discord messages shows a perma delete icon

[covert sun]

Hey @rugged flare, thanks for sharing! Could you please add a #1140622008086970420 in the channel, including the permalink to that thread.

[ancient salmon]

It feels like I should say this to the developers:
PerplexityAI has been a blessing since I started using it a month ago. Like, it is a total game-changer. The new update which will bring image and video inclusion in generative answers is highly anticipated and I hope you will deliver

Been having a blast so far
Regards

[indigo topaz]

how can I change my email ?

[visual bough]

This is currently not supported

[edgy dirge]

If I pay for the yearly subscription, is the 600 GPT4 daily quota locked in? Or is it still subjected to change.

Tried to find the answer in the tos but can't seems to find any

[crimson nymph]

afaik it can change just like how they increased it to 600 uses a day a while ago. But you can assume it will stay above 300 since that is mentioned on the website.

[edgy dirge]

Thank you

[fierce root]

I noticed that the quotes from the website which used to appear in the Sources section, aren't there anymore. I hope this feature comes back soon because the quotes were really helpful for verifying facts and making sure the sources linked to each sentence of the generated answer matched up.

[bitter grail]

Hi. Why, when I instruct Perplexity ai to generate a 1000-word blog post, it only gives me 300-500 words? The first time, I asked for 900 words, and it gave me around 320. The second time I asked for 1000 words, and it gave me 481. Have I missed something, or is there a glitch in the system?

[covert sun]

Hi @bitter grail! LLMs generally don't count words in the generated answers. You can use the following approach, prompting it to create longer or shorter texts (still it won't be the exact number of words). Here's an example: https://www.perplexity.ai/search/Please-create-an-v2hMNfRPTeyv3YD0o6hkTg?s=c

[rugged flare]

It would be helpful if, when clicking on videos presented, I was able to scroll further without needing to click on the video to expand it where the progress bar appears. Every time I click on a video from within Perplexity, I lack this feature. Having to click on a video to get it to full screen just to scroll is annoying.

[quasi cave]

what happened here?

https://www.perplexity.ai/search/Adaptive-supersampling-can-mq.jacFiRzu3okK01UX99Q?s=c

is this a bug in the frondend part? given me the result from other user's prompt?

[covert sun]

Hi, @quasi cave! Yeah, it's known a bug, we'll fix it. Thanks for reporting!

[sour acorn]

Oh my ! How did you manage to get that ?

[quasi cave]

pure luck x2 i just got it again xd

[kind jewel]

Please consider expanding to multiple languages at the same time. I would like to receive answers based on the language I wrote the question

Sorry if it has already been asked

[quasi cave]

is there an extra pre filtering when asking a follow up question? when i started a new thread using claude-2 & with copilot enabled i got this in response

[quasi cave]

but when i repeat the same query/question in the follow up. i got this instead. the llm refused to answer no matter how hard i tried to bypass it

btw 500 page is not actually 500-page long, it produces responses with ~1100 words

[true estuary]

im getting this error, im a Pro user. Help me

[worthy veldt]

its my humble request, please bring back the Discord Bot

[gray pumice]

Very informative. Lol

In my not so humble request _ where the Discord bot at??????

#backlash

If you're on Discord having bot channel is least you could do.

[true estuary]

@gray pumice What's so funny?

[gray pumice]

fact that they coded error in a way that you dont know what's going on

¯_(ツ)_/¯

like, you'd agree that some info on what really happened would be appreciated

[south ether]

Hi! please send an email to [e-mail] so we can request more information from there that will help us figure out what is going on.

[noble monolith]

can't sing in with mail, tried with several browsers. all mail option not working

[worthy veldt]

Lol I was talkin abt perplexity AI bot 🙄

[gray pumice]

i know

[worthy veldt]

Beta API support weblang results are simply interact with GPT results? actually i need results like web version

[royal inlet]

Sadly, same error for me on my laptop (phone and gaming PC are still logged in, and works fine.)
Can't sign in.

[royal inlet]

@noble monolith @true estuary Uh, I've found a workaround. Tried to connect in privatemode (on Edge), didn't get the Error message, but needed a mail confirmation, which I opened in non-private mode, and that logged me in.
Hope it worked for you.

[covert sun]

Hey @noble monolith, @true estuary, @royal inlet! Are you still facing the issue when you're not in the privatemode/incognito? Could you share any errors you see in your console when this happens. Could you please DM them to me or email to [e-mail].

[autumn skiff]

Really great service, I use it alot for research purposes and it wouldbe beneficial if u could add the date of yhe research in the output or even filter by date

[noble monolith]

turning off ublockorigin solved problem on my side, you can turn it on after you sing in

[south ether]

Correct, it seems to be due to Adblockers and CloudFlare in general. If anyone else is experiencing this please try looking for a new adblocker other than ublockorigin or Ghostery.

[near sail]

Hello Team Perplexity! What an awesome product you have... Thank you! The only thing is that my feedback on improving accessibility features in the iOS app has so far been ignored, or at least, not implemented. Is there a hope to see these changes in the future? Really, all I am asking for is headings and a more streamlined way to navigate between answers inside a long thread. Otherwise, things seem to work great! Thank you again!

[ocean leaf]

@near sail @south ether @covert sun is this normal? I only see pictures and videos in the answer.

[south ether]

Hi Vic! Accessibility is on the team's mind, and there are certainly improvements to this coming in the future.

[south ether]

Not normal, can you send me a DM or shoot a mail to [e-mail] so we can request more info?

[ocean leaf]

@south ether Okay, thanks.

[noble monolith]

@covert sun most of the time front not working, can't search anything

on iOS everything is fine, web version making problems

[visual bough]

What kind of problems are you experiencing, can you please create a #1140622008086970420 and attach any screenshots or screen recordings?

[noble monolith]

front is dead, can't search or go to any section

it's related to brave browser on ubuntu, working on firefox

[visual bough]

Yeah, you can’t use Perplexity with any content blockers

[noble monolith]

turned off ublockorigin + F5 still not working, on firefox it's working with ublockorigin activated

[visual bough]

Can you please also disable the brave built-in blockers?

Did that solve the problem? @noble monolith

[noble monolith]

disabled + F5 still not working

[visual bough]

Do you have any other extensions on? If not, please delete all website data and cookies.

[noble monolith]

have few extension that are not related to ad blocking, deleted all data and cookies. didn't resolve

[visual bough]

Have you turned off all the other extensions?

[noble monolith]

btw also working on chrome with ublockorigin turned on

[noble monolith]

working now, brave shield and ublock are both on

@visual bough

[visual bough]

Do you know what caused the issue?

[noble monolith]

no idea

smt on your side I guess

[noble monolith]

not working again.. @visual bough

[covert sun]

Hey @noble monolith! Could you make a screen rec of how it's working and DM me. We'll investigate.

[noble monolith]

it just shows start page, whatever you click it doesn't work

now working lol.. will make video if it happens again

[primal violet]

Maybe some improvment needed in wether copilot should ask these type of questions or not.

Also, it should consider the User's current date and time, instead of the server date and time. In this case, chatgpt gives the right answer. (it's 22nd Nov for me)

[whole helm]

It would be really helpful if we could attach any file apart from pdf files for example, R code file, ppt, excel, csv etc

[junior pasture]

I really like the 'User Profile' feature which globally affects the output of the model. I have two suggestions: first, could we allow users to have multiple personal settings? Because I have many requirements, such as wanting the model to play the role of a teacher, a boss, a good friend, and so on; second, switching is currently inconvenient as it requires going to the settings page. Could there be an easier way to enable or disable personal settings right from the homepage?

[wise igloo]

Well, I mean you could use the custom AI prompt in collection?

[finite river]

Is there a way to blacklist sources? For example, geeksforgeeks.

[naive oxide]

Adding parsing to detect “Copilot: <‘ask me X’>, <‘ignore geeksforgeeks in your search’>”

[vapid mason]

Unimpressed with the hard-coded antisemitic responses that somehow got put in your model. https://www.perplexity.ai/search/a786cb07-b89b-45f1-9abd-f72cc182e928?s=u what's wrong with letting your AI learn about reality, and provide real answers on global politics?

[prime lotus]

I had posted this in general, but I realized I should of posted it here instead: One thing I have noticed, and this is a recent change is that it does not always recognize a followup question to a prior response in a thread, I've noticed of late that its regarding the followups as new questions without considering the prior.

[drowsy parrot]

me too, trying to log right now with same error

[covert sun]

Hi @vapid mason! The thread you shared is not visible. Could you please press "share", so we can check it, too.

[covert sun]

Hey @drowsy parrot! Turning off adblockers or creating a rule for the site should help. Kindly check this message, as well: #👉│feedback-general message

[worn lynx]

Voice search does not support Chinese. Every time I speak in chinese. It will be transformed into some random English words

[naive oxide]

i dont think sing login is available for perplexity

or maybe try a different song, the ai cant always match from some specific songs

[quasi cave]

server down?

[slate stag]

can't log in, what happened?

[crimson nymph]

please try again with adblock disabled

[slate stag]

it doesn't work, is my account ok?

[crimson nymph]

yes it's okay. it's probably an error on your side. could you try to login in with a different browser

this error is caused by adblockers most of the time. but it could maybe be another extension or setting in your browser

[chrome rune]

It's incorrect in first sentence, then gets it right https://www.perplexity.ai/search/70fbf095-c892-4927-b5fd-abf302dac088?s=m

[pure vigil]

Perplexity doesn't use ChatGPT4 to it's fullest potential when answering questions. Here's what I mean...

I asked this question to both Perplexity (w/ ChatGPT4) and regular ChatGPT4:
"john c mcginley says when auditioning for scrubs, the sides said "john mcginley type character". i didnt know he was famous like that. what likely did the writers know of mcginley that inspired them to use his personality as a character on Scrubs?"

Here's the Perplexity answer:
https://www.perplexity.ai/search/john-c-mcginley-Sit0OQ6aSDyEQMFvH4VYYQ?s=c

"John C. McGinley was already a well-known character actor before his role in "Scrubs". He had worked with notable directors like Oliver Stone and had made a name for himself in the industry. The writers of "Scrubs" likely knew of McGinley's work and felt that his acting style and persona would be a good fit for the character of Dr. Cox. This is why the pilot script called for a "John McGinley type" for the part of Dr. Cox. Despite this, McGinley still had to audition multiple times to secure the role."

Here's the ChatGPT-4 answer:
https://poe.com/s/1hjEnzLpPOKubqawXxif

"John C. McGinley is well-known in the industry for his ability to portray intense, sarcastic, and often eccentric characters with a certain degree of relatability and depth. Before his role as Dr. Perry Cox in "Scrubs," he had played a variety of memorable characters in film and television, which likely contributed to him being recognized as a "type" in casting circles. Here are some key aspects that might have inspired the writers of "Scrubs" to use his personality as a reference:

Intensity and Sarcasm: McGinley has a knack for portraying characters that are intense, sarcastic, and often a bit prickly. This is a key aspect of Dr. Perry Cox's character in "Scrubs," who is known for his sharp wit and often harsh, sarcastic remarks.

Depth and Complexity: Despite their often gruff exteriors, McGinley's characters often have a certain degree of depth and complexity, showing vulnerability and warmth in unexpected ways. This is another key aspect of Dr. Cox's character, whose tough exterior hides a deep concern for his patients and a struggle with his own personal demons.

Physicality: McGinley is also known for his physical acting, using his body language and movements to add another layer to his performances. This physicality is often used to great effect in "Scrubs," where Dr. Cox's physical gestures and expressions often add to the humor and emotion of a scene.

Versatility: McGinley has a wide range, and has played everything from villains to comic relief. This versatility likely made him an attractive choice for a show like "Scrubs," which often balances comedy and drama.

Before "Scrubs," McGinley had memorable roles in films like "Platoon," "Wall Street," and "Office Space," and TV shows like "The West Wing." These performances likely contributed to his reputation and made him a recognizable "type" for casting directors and writers."

This happens a lot with Perplexity. It will simply use the search results as the means for answering the question. But since it's using an actual LLM, it's my opinion that it should use BOTH the LLM information and the search results. This would create better answers for Perplexity. Because ultimately, I'd love to just use Perplexity instead of having to switch AI apps all the time. I hope you guys agree and will make changes.

[undone rock]

Please increase the allowable size of the Collection AI Prompt

[brittle gulch]

Lots of error latetly when using it with coding. Its horribly and too often a waist of time. Claude and GPT does do the same errors. Any advice?
Also easy tasks like reverse a word are coming out wrong!

[left geyser]

HI, this is going to be long, and maybe coming as very blunt, but, I am comming from a good intention and wanting things to be improve.
I am a long-time perplexity user, and I really benifitted from it alot, it's my default search engine, for a quick, consensus look-up.
Being a long-time user, I have several, in fact, many critism, about it.
Perplexity, is so much hyped in the AI space, what it's doing is very much simpler processes that appear to be complex.
Searching for five or max 25 pages, and summarising it altogether, that's what I understand.
the copilot feature is very much an accelaration of the free feature and nothing more than that.
The in-between work in a typical human search like, information foraging, refining search queries, going into off-tangent searches, finding counterfactuals are not in free, nor the 20$ paid tear. To say the least, it's too expensive compared to Kagi which offer supperior experience.
Basically, humans are not a good searcher by default. The query that we type on perplexity is very limited in its density, I am not sure if perplexity work on that, to begin with.
Try to use the focus feature, such as choosing reddit and use the copilot feature, it will not obey you,

The RAG feature is very, very disappointing, even people who design basic RAG functionality would know that referrencing multiple pages and providing pg numbers are important. Peerplexity, at least in my limited testing, always only reference the book title. is better
yeah, so much hype, and that doesn't help.
please disagree, and thanks

[full cipher]

did that, still no change. the second time it listed the things i had chosen the first time when i onboarded, but not the second time.

[visual bough]

You should have access now

[full cipher]

thx. if i have feedback for the interface as well, is this the best channel?

[visual bough]

Yeah, this channel or in #🌐│feedback-web

[full cipher]

leaving it here since it applies to both.
If I enter information in https://www.perplexity.ai/profile it disappears. when i forget to edit it, i am 'living again in colorado'

which language should the ai use
As a german, i WANT the interface to be english as I am also recording training videos etc. But my output is german and english.
I will include that in my instructions / prompts but dont want that selection to interfear with the output. If it does interfear and is not just for the interface please add an other option which will result in "I'll answer in english unless told otherwise'. (similar for the app)

Also it would be helpful to have a setting for which focus I want to keep or at least for it to keep my selectin. If I am writing content I dont want to switche every single new prompt box. better yet, allow me a pompt option so I can put it into my sop / workflow.

also seconding vvoo note about multiple ai profiles. at the moment I am using extended prompt instructions to cater towards that, but it would be easier to have certain settings.
for me in a simple case: just have a german and english profile. the german profile would talk about "prefer sources from de" and f.e. the english in my case would be "prefer content from the US".

adblocking - why would it be relevant that I am running an adblocker for your site?

feature - i really like the way the midjourney discord server does regular micropolls with the users. and your onboarding to discord has to be the prettiest and smoothest I have seen so far.

[full cipher]

"YouTube - Allows for targeted searches within YouTube, offering timestamps and the option to watch videos in-platform."

does it also allow for meta searches like "here is a channel, summarize me what this channel is about"?

What is the difference between GPT-4 and Claude 2?
https://blog.perplexity.ai/technical-faq/what-is-the-difference-between-gpt-4-and-claude-2

the link at the bottom looks like a recursive link - maybe indicate that it is a shared conversation (look what we can do!)
but even if i run that prompt again, it gives the outdated information. maybe remove that? 😉

is there a way to upload files maybe to a collection? because if i want to analyze something with a variety of prompts, i dont want to upload the file each time

"Need to dig into a document? Upload it to Perplexity. Great for summarizing PDFs, explaining codes, or translating text, this feature lets our answer engine focus on your file's content. It's perfect for when you need insights based on that specific document, without external internet searches. Remember, only text files like PDFs or plain text are supported, up to 25 MB."

use case I am about to discuss today with a client: an author who has books. We are talking about her using perplexity to summarize chapters, create word lists, find ideas for new chapters ... a variety of things. for each of these prompts she needs her content. alternative: have a few slots for content available which we can reference easily in a prompt (not dropdown)

second, are there any insights from your team how to best structure the input we have?
think the above book example. i can just upload the book but i can also upload a structured summary in order to help my prompts. how do i best present my data?

second use case
export my wordpress in a specific format, have a second file (json?) with all the meta information about the page. upload that somewhere and let me reference that in prompts.

...
https://blog.perplexity.ai/technical-faq/what-does-the-ai-profile-do
links to a conversation which is not available anymore. [create a special user which only has links like this for easier maintenance)

[nimble mauve]

@cunning goblet honestly the who UX sucks man, have to do something about it.

[cunning goblet]

You can try doing this via your AI profile. Let us know if it doesn’t work

[cunning goblet]

Web or mobile? Could you provide some specific things that suck according to you?

[jovial temple]

could there be a bit added to ai profile that’s not about the profile but literally just like custom instructions?

i don’t think the whole of ai profile is put into every request and it’s kinda unclear where to put stuff like how to format output

#1178488648233144443

just sent that

[finite pollen]

@cunning goblet hello! Thanks for making the API. Been loving it. I was wondering if the OH2m7b models will be coming back. Was very very desirable. Thanks.

[kindred tinsel]

Just dropping in to say the recent changes to make answers more concise for quick questions are working well so far (for my questions at least)! Props to the team - in addition to being "less to read" it also is a lot faster since it's less to generate. And I still get nice, long, comprehensive answers for the questions that do warrant it.

[ocean leaf]

The same is here, left the server and join the server with pro invite link,still no access to pro channel.

@cunning goblet @visual bough

[visual bough]

[night palm]

I have never even heard of the pro channel

[ocean leaf]

Well, if you go perplexity pro, you'll definitely hear it and hear it a lot.

[night palm]

Are you a mod? Because i have just been granted access to the pro section

[ocean leaf]

Asking people who's in charge here let you in the fun club then.😂

[covert sun]

Hey @night palm, the team usernames are green and our community volunteers usernames, that also moderate this server, are yellow. Fixed it for you.

[night palm]

Thank you!

[pure vigil]

Oh is it tagging you gets you to respond? A bunch of people including myself left thoughtful feedback with no reply. But this guy says the UX sucks and here you are

[covert sun]

Hey @pure vigil! Tagging is not necessary, we appreciate your feedback and it goes to the corresponding tickets in case of bugs. Please file inaccurate answers you want to report them in the #1140622008086970420 channel, if those don't duplicate the existing ones. It will be easier to track them. Thank you!

[covert sun]

As for the example you shared, @pure vigil, the longer answers, formatted as a list, doesn't always mean better answers. The shorter one answers the question, too and there is also variance in answers, kindly take a look at this one: https://www.perplexity.ai/search/john-c-mcginley-qfU5o4SxSdehBdwtJXoqgQ?s=c
If that's your preferred format, you can try adding this instruction to your prompt, rewriting the answer or saving this format ("you must always answer in detail in a list format") as your AI Profile settings so it will be applied to every query you send.

[full cipher]

is there a possible instruction to force it to always create unorderd lists aka bullet points instead of ordered list? I mean i have a macro to remove the numbers, but I would like to keep the bullet points for easy copy and paste.

[lime talon]

Please add Indian banks in Perplexity Pro! Sometimes debit cards are not working..Only US Banks are present

[barren scarab]

Had suggested this related feature sometime back. Add your thoughts too if you like.
https://discord.com/channels/1047197230748151888/1155599236784730196

[full cipher]

not many people will use that but the pinned posting feature in discord is helpful (i was f.e. looking into the api general channel and the pinned posting could be "here are the 5 links to get started". low priority

[south ether]

You can use custom instructions on every collection you create, which will also help keep your instructions and threads tidy.

[south ether]

You can specify your preference for the response format on the instructions portion of the profile setting on perplexity.ai/settings, where it says "Introduce yourself".

[quasi cave]

is this good enough? https://www.perplexity.ai/search/john-c-mcginley-J6FZQ.u1RliIyPKbLyJDwQ?s=c

you have to use (all purpose meta) prompt to get this result

[lofty dawn]

I have just become aware that a new "API" tab has been added and is required for updating the payment method. I was under the impression that pro users are granted free access to all models. However, the offer of $5 per month is also quite reasonable.

[full cipher]

from the other channel - "gpt4 for short precise answer. claude for long text synthesize"

wouldnt it make sense to be able to set a model per focus as default? Basic setting = one for all. expert mode: choose one for each of the focus areas. I would asume that All / reddit / youtube would benefit from using either the perplexity model or gpt4 but writing benefits more from claude. writing also works better with copilot turned of but i have to remember that as well.

(i would also assume that using the pmodel is cheaper for your than running claude or gpt4?)

@scarlet current this ties in with your request and is also for workflow - a general choice per focus which model to use and override on collection level.

[novel onyx]

The discover tab is fantastic 🙌

[pure vigil]

what is All Purpose Meta prompt? i dont see you adding anything to the prompt? help!

[night palm]

This was on ipad

Very distorted

[violet niche]

Feature request:
Please add the choice of AI model to the main page. I want to choose model and ask question on one page - like it is done in perplexity labs

[quasi cave]

an example https://discord.com/channels/1047197230748151888/1176512713627811910

[daring viper]

Hi, I noticed that on perplexity it matters a lot if you are asked about a single product or a series of products. For example, if I ask "What is the latest firmware for Huawei Watch GT 3?" I get a correct answer, while if I ask "What are the latest firmware for Huawei Watch GT 3, Huawei gt runner, amazfit gtr 3 pro, galaxy watch 4" it gives me wrong information or can't find it- What solutions do you have for this problem?

[barren scarab]

Probably due to the number of searches involved. Out of curiosity does it work with copilot as it performs search iteratively.

[daring viper]

Yes, it would be nice if he could maybe take even a few seconds longer but still be able to answer correctly.

[urban vale]

I have a feature request. I'm not an english native speaker, so I often use perplexity in my native language. the problem is that when I ask it something it searches on the internet in that language and this is a potential limit because there is way more content in english. so basically the request is to make perplexity search the internet in english but then reply in another language

[full cipher]

make that a default instruction in your ai profile.

[urban vale]

I didn't think about it

[wild kernel]

I have the android app but I still prefer the website, is there any way to disable the 'get the app' popup? Not even with a pro account? D:

[covert sun]

Hey @night palm! Just to confirm, it was on the browser, not the apps?

[keen relic]

i just experienced something weird, sometimes it happens that one query will hang and wait on system then give up, but I have been loosing posted questions in recurring events. It's like server cache is dropping them, are there any issues right now?

dropping the sessions without loggin them first.

maybe posts are getting reject

[night palm]

It was on safari on iPad

[covert sun]

hey, @keen relic! What model did you use? And when you say some sessions dropped, do you mean the answers generation didn't finish? Were the threads you mentioned listed in your library?
Would be great if you can provide more details, #1140622008086970420, we'll check.

[full cipher]

api

[covert sun]

Hey @full cipher!) What's is the question about the API?

[covert sun]

Thanks for reporting, @night palm! This issue was fixed ✅, could you please check it one more time and let us know if you see any issues.

[fierce root]

When clicking on the "copy to clipboard" button, it would be beneficial if it provided citations not only for the text-based answer but also for the image and video search results of the query. This would group all the links together, without having to go through each result individually to get the links to the referenced media. The citations list should label and differentiate between the text, image, and video sources of the query.

[keen relic]

sorry for delay in response Alex, It's not happening today, the steps to repeat were: text input, submit post, then process hangs, gave up on analysis, and lost text input, and no history of post afterwards, I believe I was chatgpt4 but I could have had it set to Perplexity, when I saw that there was an issue at 1:20+- EST, I didn't pursue it further, if it occurs again I will capture more usable details, thanks. Is there anything we can do to enable debug on the fly, or should we just use browser inspector and see if we can catch anything?

[junior rain]

Where can I get the Pro channel invite link? Thanks.

[visual bough]

You can find the link in your settings at perplexity.ai, but you have to leave the server first 😊

[junior rain]

Thanks @visual bough I am leaving this channel now.

[visual bough]

Yeah, that's right, now you have access to the Pro area

[junior rain]

Nevermind. Sorry, I did not see the extra "Pro" areas on the left. My bad.

[lofty dawn]

If there are an option to change languages next to the "rewrite" button at all times, it would be more helpful for learning other languages. Thank you.

[pure vigil]

The Chrome plugin doesn't work:

[modern rain]

should be setting, have a space between question mark, and a period instead of a question mark

just noticed it while i was reading the faq, i love perplexity though so much <3

[barren scarab]

Probably a bit of niche ask. You can prompt to repeat something in a different language as well when you query originally?

[full cipher]

Question for @steel magnet and the team, best to be answered on your blog.

• is it only for 'search results' or can we prompt for image generation directly
• can we assume it is dalle underneath or is it your own model?
• can we create text in the output (my guess is no, just address that directly in the blog post so people dont even try it out and say buuh cant do it)
• what happens to a longer conversation, which parts does the image generation take their cue from?

it is okay if you just want to use it as a think to visualize somehting, but content creators like myself will want to know if we can direct it, f.e. to create images for the text we are crearing, marketing assets and alike. if you dont plan on going into that direction, please dont call it " generate image" as that has a meaning to it. call it "what does the machine think your ^query looks like". That is a fun feature to play around with and does not disappoint. [1]

https://www.perplexity.ai/library should indicate if i created an image for it already.

and if you have any input into the results - please understand the pouty women syndrome. i think it started with midjourney 3 or so that I started noticing that anything you prompt which is not a direct thing / concept[2] it results in a pouty, pretty, often sexy woman. It is the bane of image generation on most of the models and a reason to ask for a nsfw filter on most of the overviews. at minimum allow me in the curtesy of being able to NOT see that by something I can put in my Ai profile.

[1] i have done over 14k images in Midjourney with prompt experiments etc
[2] before you would prompt f.e. tranquility for an image and got something abstract or anything really. After that, those "i dont know what to draw / paint here" became "something pretty women"

[lofty dawn]

Certainly, it would be better if there is a button that could achieve the same effect, as I wouldn't want to write prompts myself and contaminate my questions in input box, or affect the coherence of the context.

[full cipher]

no it is not. the learning language part, yes, but i have a text expander macro for "do it again but in german". The ais work better when you prompt them in english, but then create the output in the other language. true polyglots are rare, I am more of the typical example - i write mostly in english, but my content needs are english and german, depending on the context.

[full cipher]

i can recommend the browser extension textblaze to you. https://blaze.today/ it is free and you can create a snippet that will react when you do f.e. /l for language. Create the prompt you wish to add at that point. you can also add to it "find interesting vocabulary in there" or other things you are curious about.

[fierce root]

This happened to me too. I was able to summarize the article via the website interface instead but had to convert the webpage to a PDF file. Hope this issue gets fixed soon in the Chrome extension.
https://www.perplexity.ai/search/f354c698-7caa-485e-8dea-baae9d93aa55?s=u

[haughty bloom]

Web app says use the iOS app because it is “powerful” yet web app has more functionality than iOS app (no way to set focus or upload file in iOS app).

Also, there is no way to edit previous prompts anywhere.

[harsh trellis]

I'd really like to fix an AI model to each chat. It seems like if I create a chat bot, then share that - for example with Claude, then if I later change my general Perplexity AI to ChatGPT, that previously shared chat will change its model too. I want a chat to have the model I selected for it, not change.

[full cipher]

or fix at at minimum at the collection level since f.e. gpt4 and claude have different strenghts

[harsh trellis]

Yes I'm surprised you can't do this. I'm comparing to Poe, that I'm also using on a trial, and it seems Poe is more flexible in that way. Although Perplexity has a much nicer interface

[full cipher]

i 'fix' this by currently only using claude - and direct chatgpt f.e. for coding. using pplx and gpt4 for coding request did give less stellar results (whats the word - worse?) than going free chatgpt 3.5

[covert sun]

Hey @haughty bloom! Focused search is available in the apps - you can choose it in the top right corner. The file upload function will be added in the future releases.

[haughty bloom]

Thanks Alex

[covert sun]

https://twitter.com/GillVerd/status/1731111241261957208

[mild glen]

if he's cool it's cool, thanks was just surprised wasn't tracking he'd okayed 🙂

will delete, should've DMed instead of public posting

[mellow bobcat]

https://www.perplexity.ai/search/a832a90c-62c4-4803-8367-ef182fa6767b?s=m would be nice to see the results show up in the same language as the query and sources 🤷🏾‍♂️

[covert sun]

Hey, @mellow bobcat! Did you try with Copilot? You could also set Telugu as default in your Profile settings.
https://www.perplexity.ai/search/2507661b-fd9a-44fb-b16e-6f3f0df8d02e?s=m
I got this answer.

[full cipher]

not sure what my feedback on this is, so just fyi -
through a reddit post i got to the question of
how to make a osm travel map.
https://www.perplexity.ai/search/somebody-on-the-j5RSeH3cQp.XMEB95mFZmw?s=c

the resulting links included a 404. (the osm plugin). not sure how you collect the data, but the site looked dead for a while.

[covert sun]

Thanks for reporting this, @full cipher! I'll add this Claude thread to a ticket.

[harsh trellis]

I'm discovering that the IOS app is missing a lot of features - ability to upload files to a chat, inability to set focus. Will these be added soon? It's a very competitive space! Thanks

[visual bough]

#🍏│feedback-ios message

But the focus option is already there at the top right 🙂

[sacred schooner]

When I begin a thread using a specific model I would like that model to persist within that thread even if I switch my default model in general settings. I get more value out of Perplexity when I can effectively and easily use Claude 2.1, GPT 4, and Perplexity models without having to constantly go into account settings to swap between models.

An alternative would be to allow model selection from within a thread, or even during prompt input, instead of at the account level. Thank you.

[pulsar ravine]

When I'm in my thread collection and I delete a thread, the dialogue box doesn't go away after clicking confirm. Win10 chrome.

Edit: refreshing the page is the only way to fix this and make the collection update its view

[visual bough]

Are you using any extensions? Maybe content blockers?
If so, please deactivate them and try again!

[pulsar ravine]

I'm unsure if I need to relaunch chrome. I tried disabling all the extensions but I get the same behavior. I will update you when I relaunch it

[crimson nymph]

#1180215817405734932 message its a bug they’re aware of it

[harsh trellis]

You’re correct! It’s well hidden though…

[scenic oak]

During a conversation, it seems like it loses context when I edit a message I wrote, even without changing the model (which I expect to retain the whole context even if I changed models)

Also

Quite some LATEX is not getting rendered, here's an example

This is quite an annoying problem

[crimson nymph]

Do you have copilot enabled? And for the latex problem please create a thread in #1140622008086970420 and add your thread where this problem is occurring this will help the team to find the problem.

[scenic oak]

No, do I need to? If yes, why?

[scenic oak]

.

[crimson nymph]

There is no need to enable it. I know there sometimes are people having some issues with context when it is enabled in some rare cases.

[scenic oak]

Should I report?

[crimson nymph]

Yes you could report the problems you’re facing by creating a thread in #1140622008086970420 or click on the flag icon under the answer on the webpage. This will help the ai team to improve the output

[woeful igloo]

Perplexity responded with a numbered list of answers to my previous query, and I asked pplx to “expand on 7” - to mean the 7th numbered item on AWS. Copilot got it right, and performed the right searches, on AWS.

However, it responded with info about the numerical number 7 in math; instead of the topic in the 7th numbered point and the correct searches. Is this a known issue?

Perplexity’s response:
“The request to "expand on 7" is quite broad and could refer to a variety of topics. However, given your your interest in tech, I'll assume you're asking about the number 7 in a mathematical or technological context.
In mathematics, 7 is a prime number, which means it can only be divided by 1 and itself without leaving a remainder. It's also the fourth prime number, following 2, 3, and 5.”

[void moon]

I got this problem, too

Even without me switching modes, it has turned into online search mode, no longer connecting with the context. I had it in Writing mode before.

Without any warning, it switched automatically and can't be switched back, the distinction of the previous content is lost.

[nimble tiger]

came here to tell, that your service have stunning UI/UX!