#general
1 messages · Page 149 of 1
It is amazing truly
I love the game
You can improve towers, overclock them
grok 5 ?
also what stops you to give this game more update
where the game?
You buy towers, you put them on the map
Click start wave and you have to defend your core
When you click on your towers, you can improve them
There are bosses also
Goal is to avoid ennemies going there
Omg
so like Bloons Tower Defense?
Click on one tower and improve it
did you know you can still update that game
Don't know the game but probably!!
Try it
This aint good
.
Its shooting at nothing
What?
Its so broken
Click on Start wave
oh okay good to see here who isn't a vibe-coder but yes this is impressive tbh
It is not lmao
its working for me wdym
Reboot doesn't work
Only for you then it works good for me
I've been playing this for like 30 minutes lmaooo
Isn't it good?!
Oh yeah reboot doesn't work indeed
But just replay
like refresh
it's not because it's badly coded
it's because of codepen
the gemini 3 made this game in a just one message?
Can you update it, the game goes too fast
And cant get trough wave 2
Yes
I am at wave 14
Click on your towers
and you can inject them to make them more powerful
yeah you are doomed
Mortar is too slow and has no damage
I tried with only gatling full power
a lot of them
Still died
skill issue xddd
real
brotato chop has never played a single tower defense game 😭
brooooooooo wthhhhh
steam?? this simbol
hell nah bro
I don't want to be banned
Black balls too op lmao
wait until you get
Tesla
it's super easy to beat them
what?
serious you don't seeing something ? 🤣
what
i'm so confused
can I dm?
oh i see it now nevermind
youre the one who thinks about that
yeah but
its unrelated
lmaoo
but yeah
funny
you know
but if pineapple see the same thing we are cooked 😭
nahhh
bro it's an ai generated game
there is nothing weird lol
Trying a new game now!
I think the cuphead is soo much for gemini 3.0
Hello, I'm here for VideoArena
this game is hard.. but super impressive
Okay I'll try
It made this lol
It's just a boss battle
I saw it. You're in trouble now. @jovial sapphire too.
Noooooo!!!! 😂
What do you think about the game
Well, I think that's pretty impressive. You guys are good. Nice jobs in there.
better that I thought
The thing is, all of this is one shot
If you provide it with textures
.png
This community is dead
Nah it's not
they were never born
The dev mode "legit" community is the real lm arena community 😶
Someone can share the url for using lithiumflow?
damn wat happened?
Add image to have a lot more or LithiumFlow
tysm!
@fiery gull @jovial sapphire created "people playground" physics sandbox in 1 shot single html file with lithium flow
try it out yall
lol
It did a cuphead style game and it's quite funny
Nailed it on the cup head charachter!
hi mans can someone tell me in DM how can i create scary story for tiktok??
And how can i create prompt profesionaly i have google gemini pro
?!
nah
fling the ragdolls on the walls
did you select PC
Perplexity is offering perplexity Pro and earns on each referral
You Guys Can Easily Make 100$ to 1000$ with this Perplexity referral
- No drama
- No lies
- No clickbait
100% Real Earnings
Claim offer here and download perplexity Comet for free through this link: https://pplx.ai/muhammadze84187
:/
hello
go to sora 2 website
and look at most recently generated
the quality is severely degraded from when it first came out
its not terrible but its by no means as realistic as launch
the physics stand out the most
@echo aurora hi
i have a question
when i hit this
retry button
is it the same model that is used?

Yeah, I believe so
okay ty!
@echo aurora

like tell me this looks real at all
lol
come dms @tulip tree
Ok
sent you a code
in fact
tell me if its working here plz
she climbs THROUGH the barrel of the tank turret
if you watch closely
like phases literally through it
lol
@stray aspen
LOL
10/10 physics
did the code work?
the click mouse don't work ;-;
awesome work!
-# ||The purpose of life is: to have fun, play games, create games, share games, discover awesome stuff ^^ ||
2026 shapes up to become the most awesome year we ever had
AI will probably also become cheaper because of increased competition.
damn, my new notebook is soo good
I need to talk it to stranges :3
sora 2 don't have so many vote to say if is 1 or 2 or 3, but the lmarena talk 1 🙃
I try to defend lmarena but is hard lol
they aren't even making any money now, how would most of these companies surivive price cuts?
AI ever is an investor gap
I think nobody LLM make real profits
easy: as soon as we have AGI, we enslave robots to do our work so that we have a post-scarcity economy :)
in 5 years it might happen, right?
a post scarcity economy is a lot scarier than it sounds paws
Or website makers
so truee
a futuristic StarTrek economy doesn't need money
ok, it might take a little longer than 5 years..
because there's this issue with the climate and stuff (migration, crop failures, etc)
Our future is gonna be so good
can you PLEASE for the love of god change it so it doesn't pick image generation automatically when attaching image🥀
who decided this was a good idea
if we manage to get an aligned AGI, yeah
(that's quite a big if)
put that in feature requests
Fr
I hate that so much
Could happen really fast
with an algorithmic breakthrough, maybe
NEW HERE TO LEARN THANKS FOR ACCEPTING MY INVITE
so your definition of agi include the idea that is conscious or conscious adjacent?
an AGI is, if put in a suitable robot chassis (android), able to do (almost) everything a human can
This is a weird definition of agi
so like for example, drive a car?
it can do everything, except of the biological stuff (eg. eating/breathing/drinking is not necessary)
yeah that definiton could only really happen with continual learning
also like if you want to get robots to do most tasks, there's a lot more efficient ways to do that then make a single robot to do those tasks
yeah, specialized robots are often better
but, to found an off-world colony for us, androids are a good form-factor
does this mean gemini 3 wed?
twitter analysis is the worst
uh, not to be a drag, but thats not gonna happen any time soon
the off world colony thing
that means gemini 3 boat acutally
doubt
and after 2030, on the moon
yeah i think its coming this week 🙂
actually, i think it'll happen (much) earlier
yeah thats the main problem with this
2200 might be too late, because of accelerating climate change :/
(Venus hothouse scenario long-term)
...
there's zero universes where colonizing mars is easier then fixing climate change
but hey, ai is accelerating climate change so maybe its the exact opposite of a solution
yeah we could literally spend a few hundred billion a year to keep temps extremely stable through lower stratosphere aerosol injection
like totally tractable problem
imo
or.. we build arrays of mirrors
on the ground/ocean
tysm! its a very fun game
it's the first time ai made a game that was enjoyable
i spent more than 30m on it
or beat a videogame that doesn't exist in its training set. Models now can't even beat 10% of a very well known game.
to accelerate climate change?
How many prompts did it take
no to reflect solar energy back into space (https://www.meer.org/)
Join MEER in restoring Earth's energy balance. Explore our climate change solutions and eco-friendly cooling systems. Create a sustainable future.
one
yeah but you get multiples in terms of radiative forcing / $ from sulphur
it's one shot
a hundred few billion per year is a lot of money
on webdev arena?
very useful
and cost curves would likely be pretty easy to move along
What was the prompt
my point is that it doesn't (relatively) cost that much to save the world if we discovered the world was in an Earth-ending positive feedback loop
maybe increased water vaporization leads to more clouds worldwide, which also increases the reflection, so that we have a cooling effect?
"global dimming effect"
find me a few hundred billion per year and we are talking
im saying we'd find it if the world was actually in a death spiral positive feedback loop
like trivially really
you have a weird definition of trivially
wdym
It's actually a negative feedback loop (I forgot the name, but it's an actual thing in ecology). Negative as in it spirals out of control due to outgassing of methane.
that's positive feedback
find me a few hundred billion per year right now that I can spend on fixing the planet
you can't because the world is not actually ending
we're going to see like slightly higher costs of food, disaster spending, etc than a counterfactual world
scratch that really
we are spending like over a trillion a year
from what I've heard, that would cost even more money than the other guy's idea
?
on what
various climate initiatives
I was actually the question asker 8 years ago lol, before I deleted stackechange account
even more if you count lost growth from a variety of environmental regulation
I tried the 1 black pixel upload trick and gpt-5 chat hallucinated an entire image ehehe
not saying its bad, but we clearly are spending significantly on climate + environment
wow you've grown so much
1 source, 2 if that was the case, why wouldn't we just allocate some of that 1 trill to do the other plan if that works
Yeah, basically, as seen in ice core samples, the greenhouse gases move together
pretty much because people find stratospheric aerosol injection icky
When you increase one, it leads to a feedback loop that causes an increase in all the others
(water vapor is a greenhouse gas too)
24 June, London—According to Climate Policy Initiative’s latest ‘Global Landscape of Climate Finance 2025’ report, annual climate finance reached a record-breaking USD 1.9 trillion in 2023—more than doubling over six years. Between 2021 and 2023, climate finance grew by an average of 26% per year, a substantial increase from the 8% a...
there is another way: weathering
you would spray stone dust on fields which would bind CO2, as i understand it
but you need huge areas to have an effect
well I'm glad they are using ai to write their articles
the cool thing about SAI is you basically use the same radiative forcing multiplier principle that causes climate change in the first place, it's around 1 gram of SO2 -> 1 ton of CO2
office workers love slopifying their articles
most people don't notice it haha
well I do and I hate it
1 gram SO2 leads to 1 ton of CO2? That's pretty crazy
Actually hard to visualize a ton of CO2
Because it's usually gaseous
yeah the cool thing about stratospheric SO2 is you need so little that you wouldn't get appreciable acidification
and it has a residence time of about a year
its a half life really but you get the point
I just checked the report and most of the money was spent on making more effiecent energy systems or making new energy systems like solar panels, which isn't really the same thing.
If I buy a solar panel to run a server farm by no means did I decrease co2 levels and by no means did I do that for climate reasons
agreed, but you must admit that the opportunity costs we are imposing on ourselves are higher than 200 billion?
especially if you include environmental regulation generally
yeah
enviromental regulation generally saves more money than it loses
huh just had mistral of all models do better than orionmist
the orionmist output was far too terse and offered no real feedback
at what task
yep exact same thing happened to me, way too short
I posted a series bible for a story idea
That's an interesting trade-off 🤔
If it works ig
yeah I wouldn't reccommend either orionmist or lithium for story reasons
lithiumflow isn't bad for story stuff, I wouldn't pick it over gemini 2.5 pro though
They focus on coding mainly
At least that's what the improvements show
so if its gemini 3.0 you'd pick 2.5 pro instead
yep
yes
which is why I don't think it's gemini 3.0
or if it is, it's gemini 3.0 flash or something
that or google decided it doesn't care about creative writing
that'd be sad
which version/size of mistral?
didn't check tbh, and I already deleted the chat
So basically aligns with phantom experience?
That would be the optimistic outlook ig
what about the others?
deepseek, GLM, Qwen for roleplaying/creative writing
and Kimi K2, and the new Ocean AI models
Well DeepSeek is basically trained on Gemini 2.5 and ChatGPT, just by word distributions
And GLM on Gemini 2.5 Pro
And Kimi on o3
and Ernie?
I was using deepseek for creative writing until I realized gemini 2.5 pro was way better
I started to get disillusioned with deepseek around 3.2
mmm
Not sure
It's kind of funny, because GLM is closer to Gemini 2.5 Pro than Orion/Lithium is when you look at the word distributions
which tool do you use for that?
Slop forensics by Sam Peach
But modified slightly for smaller sampler sizes from manual inputs
He uses Jaccard similarity as the similarity measure over the word frequencies and ngrams
wow, sounds pretty scientific
what was the longest RPG you ever had with gemini 2.5 pro?
(if i may ask)
RPG?
Ah idk I use it for explaining STEM concepts and for large context
science/technology/engineering/mathematics?
I'm not sure I'd use it for writing, it's pretty easy to tell it's Gemini, at least from simple analysis methods
(in europe this abbreviation is not as common)
Ah yeah
Interesting
*continental europe
so gemini 2.5 pro is currently the best model for RPG/creative writing?
but mistral is second now?
It's funny because Europe and Asia are on the same continental plate
??
it seems so, but the indian plate is crushing forward into it
IMO yes
Idk, I don't think it's at the top of EQBench for creative writing
mistral is definitely not second best though
I don't really have the mental power to test LLMs at creative writing... Too much reading xD
Which is why I'm surprised that creative writing has a high discrimination factor
what about qwen?
Hi Pros, I am curious if there is a way to generate videos without anyone voting on it. I'm really an amateur at this that's why I want it to keep it private
To me it's much easier to tell models apart with technical stuff
wdym
of course its easier to tell the diff with creative writing
Is it? Not really...
is it (Mistral) best?
use grok image or get a sora account
That's why RLVR (reinforcement learning with verifiable rewards) uses maths/coding
(to me at least)
is this image to video ? I'm really sorry, i might have a couple of questions if you dont mind
most ai companies are focusing on stuff like programming, which means they are gonna be more optimised for that and devations from "what is good" are gonna be corrected, there's not even a definite good in creative writing and because its not as important ai's will naturally devate
Or do you mean tell the diff like distinguishing which model is which?
I don't know where I'd rank Mistral but I doubt it'd make my top 5
grok's is
yeah
Oh I meant more in terms of judging intelligence
where exactly I can't find it. I'm really sorry for noob questions.
That's a good point too... easier to benchmax on maths/code
ah ok
has unlimited
Grok is a free AI assistant designed by xAI to maximize truth and objectivity. Grok offers real-time search, image generation, trend analysis, and more.
but not the best
probably will improve over time tho
Ig it's good for stuff like common sense
Thank you, I'll check it right now. While Haiku 4.5 is currently down for me
I had a few tests where I asked it to complete writing examples where I inserted a bizarre situation to see if the model could detect it when I told it to finish the paragraph
I don't think there's a great answer for this yet, you might want to try figuring out yourself
Which do you think will be better?
7
10
1
Gemini 3
tho I do know claude isn't the one you are looking for
4.5 recently scored at the top of EQBench's creative writing benchmark
Both long and short form
rlly?
Yeah
sunsweeper
will opus-4.5 have much longer context?
do you have any other hard prompts?
That's the benchmark with the model similarities dendrogram
oh I see why
Yeah I'm going to test one out where only GPT-5 High gets it right
Was from a phd pharmacology student (I think?)
But it shouldn't be that hard for someone in that field
@sullen quest Hi Sir, I found grok and looks interesting. I'm also curious about Sora. How do I try it?
Just that all models failed to identify a fatal combination of drugs 💀
that bench uses sonnet 4 as a judge
GPT-5 High got it as a hint
well you need a sora 2 code, I don't have one, try going on openAI's discord and asking for one
Yeah but they test other judges too (Judgemark), can see how they scored Claude (usually order is same for top judges)
can you send me it?
i'll try it on gemini 3
I see, Thank you for the info. ❤️
AI judges… that reminds me of Buck Rogers in the 25th century with a certain "dr. Theopolis" (https://en.wikipedia.org/wiki/Buck_Rogers_in_the_25th_Century_(TV_series))
an AI, the only one which believed Buck to come from the past
and Twiggy (was a fun series)
okay!!
telll uss
meh, doesn't solve the ultimate problem, you know what I have to say about llm's when it comes to judging llm's, they tend to prefer their own
what if they used a triumvirate of AI judges?
like gemini, GPT and claude (and/or grok)
@echo aurora
DMed
Ok will DM
Just to prevent it being on the web = in training data lol
woah\

I think human is the only way to get it reliable
you don't know or you wont say?
is slower
both
He won't say
LIES
Ig it already is since it's in the dataset :/
yeah, but its necessary to be reliable
it's the truth!!
Hi , I keep getting failed to send message. Why?
when I have a VPN on it happens to me
Slightly, but you can see how the rankings by other top LLMs (on older data I think) are mostly in alignment
This can happen for different reasons sorry to say. Most commonly is you're being rate limitted and will need to wait a bit. However, I've seen cases where: refreshing the page, and trying a new browser can help get past this. And yes like already mentioned turning vpn on/off sometimes helps as well.
well I'll take it I guess, 4.5 is better at writing than I thought
Got @jovial sapphire to try the pharma question without the hint, and looks like Lithiumflow failed to get it right (without hint so far)
ok, try again but specify it is a really hard and confusing question as a hint mebe?
(Definitely don't use LLMs for meds lol, since it fails to identify the fatal drug combination)
lol
will keep that in mind when recommending substances to use to doctors
It's just something that someone in the field would know by considering the mechanisms, but that combination is not common (i.e., not in training data)
The veo glaze is crazy ngl
I gotchu
idk most people here seem to not like it as much
How?
Idk why it's ranked #1 tho
With the hint it got an even wrong-er answer
lol
gemini 2.5 pro isnt answering for me
Lmarena SITE bug
The lmarena is incredibly buggy now, it's a plan of the owner of AIs to we cant use the AIs
I want to melt glaciers with opus 4.1 for free 😡
We certainly have ongoing bugs with the generic "Something went wrong" error message + models sometimes getting stuck, but I hope as of late things haven't been noticeably different?
its had an uptick I think
its the only time i nnotice this every ai works its just that no matter what gemini wont respond even in new chat
Gemini 3 pro will …
9
15
4
be good, for most purposes
hmm how recent?
There are still people that report the issues listed above; however, the volume hasn't appeared to be anything out of the usual.
is the latex formatting issue fixed?
I've seen a lot today, also some models yestorday took wayy longer than usual before starting
arena biased towards google models?
i hope not
like there's no way the old veo 3 is ranked above these recent models
How much it costs, thousands of promts to claude, and other ais?
and how has 2.5 pro always been at top for a long time
veo 3 had quite a lead at the start
Yes its google they have a lot of data
hard to tell since some of the api access lmarena gets for free in exchange for testing reasons, I have no idea how to factor in for that
the 2.5 pro that was originally released isn't the same as the current version
Why they don't share the updates
a long time ago every time 3.5 was changed openAI would list the changed model with a different number
at some point I guess companies stopped doing that
What do y’all reckon the best ai app is on ios?
Gemini is pretty bad imo, same for ChatGPT
AiStudio in web
2.5 pro 100% free and 100% of free to control the 2.5 pro
claude sonnet 4.5 thinking for me tbh
nothing else comes close rn
But the AiStudio is 100% free
Why you wearing bonnet
he said best not free
...
: D
Flash 2.5 lobomitezed?
thats not flash 2.5 thats pro
: )
:::)
Tomorrow is a special day for the AI Studio team.
Since May, we have been heads down building a brand new AI vibe coding experience to accelerate the path from prompt to production with Gemini.
Can’t wait to show you all :)
so we finally have it
aye!!!!
gemini 3?
read the twitter post
ok 😄
cause that means gem 3 will prob be baised off of this
:0
is claude sonnet 4.5 thinking even good?
compared to gemini 2.5 pro
.... its in a different league than 2.5 pro
2.5 pro is pretty bad latsly
really? in what sense
Try claude and youll see
Anything is better than it rn but sonnet 4.5 is the best overall imo rn
Yeah but ive never hit it
oh wrd? how frequently do you use it
Uhhhhhhh
The next tweet
"Tomorrow is just step 1, the next 2 months are going to be a lot of fun."
Gemini 3 not for another 2 months confirmed

I'm guessing orionmist and lithiumflow are gonna end up being Gemini coding models
Haha
Gemini wont release another ultra?
I like how they call it vibe coding experience
That's good if true
Apps are pretty bad
Idk why but even Discord seems buggy
Messages lag and load on batches, etc
Perplexity iOS app doesn't even work last time I tried
Networking seems bugged on iOS
Scuffed phone, I suppose.
Can I share a link to an external page?
depends on the external page
whats the page
@sullen quest DM?
no
what the hell are these
The left one is Microsoft's new AI model, while the right one is an anonymous model (what I can only assume is either GPT or Gemini.
Chat, I have news regarding Gemini 3.0.
So, if we end up getting the X28 model (the one known for its out-of-this-world coding capabilities, then we will be able to make stuff.
model's namecode?
hello
Hi Pros. Does Grok website have limits? or is there a trick.
Anyone know if there is an api to get llm usage via lmarena? I want to make an iOS app version for lmarena
I get 5 times a day which is awesome. Today i'm two for 4 that have had no audio. Anybody know why?
@burnt cypress Please head to #1397655624103493813 for a detailed guide on how to use the bot
@burnt cypress You can generate a video based on your prompt, generate an image based on your prompt or generate a video based on an uploaded image and your prompt. All the information is located in the #1397655624103493813 section
what model are you using
nevermind
What is this benchmark? Amd how respected/good is it?
Apparently, it's called “KingBench”, and judging by its likes, it seems pretty respected.
This sounds like something else and not a 3.0 announcement.
But numbers are much higher compared to Sonnet 4.5. makes me doubt it
Well, perhaps that's just how good it really may be.
It would be surprising if thats the case . But 🤞
Hello all guys how i use text to video vo3 kindly i am new kindly tell me plz
On website lmaarena i cannot show vo3 or sora 2 pro
Please read #1397655624103493813 for more information. Please note that you can not select models, they are randomly selected for you. You have 5 uses per day and 2 videos are generated in each round so it's likely you will get Sora 2 Pro or Veo 3.1 at least once.
can anybody help me with the seedgram 4 high res model, as when i try to edit the images, it works fine for the first 5-6 images, but then it just shows the "Something went wrong while generating the response. Please try again", for the rest of the images i send to the model.
hello
Hello everyone and thank you!
hello
I'm seeing this since yesterday, could you please have a look there @echo aurora ? thx a lot
hi
which model is this?
guys
I got a few complimentary Perplexity Pro invites (it includes access to ChatGPT 5, Gemini 2.5 Pro, Claude, Grok, and early access to all the new models for free).
If anyone actually wants to try it out, just DM me and I’ll send over the official invitation. I just don’t want the invites to go to waste since they’re limited so please don't ask if your not going to use it.
this place needs serious moderation
alot of this foreign stuff, we don’t understand the culture, or the language, it maybe actually promoting some harmful and dangerous indirect racial ideologies or even worse, promoting propaganda and misinformation without even understanding what the hell some of these videos are even about.
oh yeah what the hell
please
there is so much slop being generated in the video thing its unreal
dude it’s nothing compared to in the wild
yeah i know but its mostly people coming here to use the top models for free
the whole Internet is just being bombarded with slob. It’s ridiculous.
like they need to have some sort of alt account prevention
half the accounts generating stuff are just made to join this server
in my opinion, that’s all irrelevant minor headaches. The concern is the misinformation that we may not be able to catch because we don’t understand the culture language.
we can imagine this place being a source of some kind of propaganda or something without even realizing it how crazy that would be
you got people from Hindu India you got Muslims from Pakistan and other parts of the world and I don’t mean no disrespect or I don’t mean the wrong way but we know that there’s conflict between some of these groups and individuals
like there are here in the United States with Republicans and Democrat. It’s no different.
except we don’t have this kind of secretion violence that some of these other nations do
Sonnet 4.5 humiliates the 2.5 pro to be honest, but the AiStudio app itself is very good, in the app Claude only has the free haiku 4.5 available
and the thing is 95% of it is probably innocent and no harm and just regular people using it for regular means
its a huge problem in the ai industry not just here. Extreamly poor gaurdrails to actually prevent real harmful stuff and it’s just very irresponsible on all these platforms that can’t even comprehend exactly how all this translates to in the real world for all the wrong reasons
once something is generated, you can’t undo it lol
I still don't understand how people can call fake videos and chatbot text harmful
Is it going to jump out of my screen and stab me
hello
real harmful stuff? I wouldnt say the guardrails are extremely poor, to the contrary, those guardrails are over the top and causing way too much false positive that could be more harmful than being useful, it's a problem of context reading, even humans are bad at that, how do you expect machines to do it more accurately?
@twin ferry @pale scaffold @pseudo schooner @wintry depot Please head to #1397655624103493813 for a detailed guide on how to use the bot
@shrewd gyro Please head to #1397655624103493813 for a detailed guide on how to use the bot
Thank you. This has been removed.
Content was removed, thank you
hello
Pinged or perhaps even dive bombed? 😸
As ChatPGT is noth used and discussed here, I add this which I found interesting.
'They cannot do world building = cannot write an actual story. They cannot play chess, they only do illegal moves. There's no actual reasoning or understanding behind the behaviour - they just follow the rules set up by the training.'
https://youtu.be/emHCav2pxLA
Our documentaries:
https://www.youtube.com/watch?v=7_j9KUNEvXY&t=
https://www.youtube.com/watch?v=BzAMjU14z4w&t=
https://www.youtube.com/watch?v=ntnynBDbmaQ&t=
The latest version of ChatGPT is ....“overdue, overhyped, and underwhelming”
That comes from a new New Yorker piece by Cal Newport, an MIT-trained comp...
I'm from Kurdistan، where are you from؟
Hello from the Philippines
Qwen 30b instruct won against grok 4 in a coding task lolool
is it only me, Haiku 4.5 is down?
hello from Korea
North america
Does Claude 4.5 sonnet have vision
not the right channel
hello
Is hunyuan 3.0 working?
This channel need some serious channeling to exorsise these prompting guys who cannot read instructions.
agreed
I'm in Iraq in Kurdistan but Kurdistan is not Arab and we don't have Arabic language we speak Kurdish in which country
In what country.
Canada
Tbf the internet has always been mostly slop
It's just thanks to search engines that we see good results
Well it's even more than what it used to be, from AI generated Youtube films supposedly to tell something of interest - yeah even I have taken a wrong turn more than once.
I think that's the fault of the YT algorithm
AI voice, AI chosen clips - which sometimes not even correspond to what the voice is telling - in a few cases even having add a scene that is completely off topic.
They removed downvotes, for example
Indeed, well I left YT and now refining my films even more adding actual animation for several scenes to have both the looks and storytelling improved.
So I don't mind the challenge.
I don't know why they stopped recommending me science channels that I'm subscribed to
Instead below each video it's all these random AI-narrated movie shorts
We're on the same page there, I still watch YT - and it's 95% science.
It's a nice country, but I think it's mostly freezing.
In what country.
Australia
It's a wonderful country.
I'm in Kurdistan.
I haven't gotten recommended a 3B1B video in years, yet I'm subscribed to him
I'm sorry we're talking about a private chat.
And where did Numberphile or Computerphile go... they're publishing videos but it never shows up
Have a look at "Looking Glass Universe" channel, she demonstrate quantum computig in a very good way.
In what country.
You're right, have not seen any recommendation for numberphile either.
I just checked, they published a video 5 days ago
But I literally have to try and remember the good channels, they're just not showing up
this is not a private chat this is the general text channel, everyone can see
You can talk to me on your own, and I sent her a friend letter.
In what country.
Accepted
I noted this quite some time backm and their terrible algorithm was one added reason I decided not to use that platform.
I feel like someone should write a better algorithm
(there probably is, haven't checked)
Long overdue, with my interest for science - I see recommendation for 'Atlas is alien spaceship coming to kill us all' - or something similar over and over. 😹
I need an app to manage my subscriptions + playlist anyways. Should write one.
Poor Anton Petrov tried to set things right in his last video.
Oh I have notifications from him turned on so at least he shows up
Facebook seems better, at least Veritasium videos show up there
Although FB's algorithm is kinda scary. If you click on too many cat videos, your feed will be just cat videos. It becomes boring.
Veritasium show for me also - sometimes.
I've made a complete cansel on FB, not there - and wiped the trace I ever been there as much as possible.
I set up daily science/tech notifications on Perplexity
Seems a bit odd today, it listed GPT-5's release as the top AI news item
Sounds good, I'm a researcher in real - most I'm keeping track on for work is stuf come to me faster on other channels.
But it's interesting to see how these guys comment on those findings some week later.
Perplexity's news feed is oddly good
That's why I mentioned Petrov specially, as he generally get things right.
Whatever Perplexity's news algorithm is, it's good
This is literally the first time I'm hearing about this
Unable to generate pictures on hunyuan 3.0 . Is the server down?
Indeed, they're trying to get AI into my phone, even Opera browser had a pop up yesterdy about it.
...nedless to say, I'm not letting that happen, neither in the computer = not going Windows 11.
AI is nice for many of my video clips, else I would be animating myself to death - but I draw a line after that. [Music scores are non-AI.]
I suppose it's like Github Agents (idk, haven't read it)
I'm reading this rn: https://kyutai.org/next/codec-explainer
"Neural audio codecs: how to get audio into LLMs"
Top article on Hacker News
A new promotion by perplexity to get 1 month pro for free if you download comet browser and made your first search https://pplx.ai/amanxy
most likely Nov Ist week 🙁
waiting ...waiting ... waiting..
Perhaps.
But, we always have to be hopeful.
HI. i am new to this server
Bro, I downloaded it, installed it, and did some research... how do I redeem the subscription?
It get activated instantly just check your profile
still asking to upgrade to pro
"Gemini 3 sooner but not this month" said by a google deepmind on september 💀💀💀☠️☠️☠️☠️ I still donno what does mean sooner for him ?
please anyone claim
it
because
it give
you 1 month free pro
lie, it doesn't work
also I will earn 2 bucks
Wait
a minute
Really
you dont get nothing
you can send a email to them.
and you chat to perplexity or not
Try to chat
does you claim invitation
on page
hello everyone, i saw a youtube video about free video generator. so i definitely got excited. i recently created a horror youtube channel
you are trying to deceive everyone
OpenAI browser coming out today
They posted on X.
Can i make videos more than 12 seconds in here?
their marketing is much better than the product itself
2 BUCKS PER INVITE?
Hello, I am Maria, exploring AI
for US it is up to 15 and France 10
and even For one moth you can also refer and earn.
moth = month
for clarification
DeepSeek-R1 i guess
hi
anyone there??
can we make video in the ration we want bcz there is not any option for slecting ratio
hello - sorry to say you're unable to set a specific ratio when using Video Arena
hello all
hello 
thanks for the info mate
Since you don't offer bounties for vulnerability disclosure
Could you consider creating a hall of fame?
how can i find the videos that i created ? the list goes really fast
that's an interesting idea. As in here in the Discord on on the site itself?
Website
Would be useful for recognition
Like put it to put it on the resume
You'll be pinged when the generation is ready, you'll also receive a DM from the bot that'll link you to the generation.
Otherwise there is nothing to gain to report
i know but since the window is closed now. is there any way to access your videos again ?
yeah that could be a nice recognition. I like the idea, thank you for sharing.
No problem! Here is an example: https://www.codingame.com/work/vulnerability-disclosure-policy/
Vulnerability Disclosure Policy (Updated November 2024) Codingame intends to keep its service secure and protect its customers’ data. If you’ve found a security issue that you believe we should know about, you can submit it to our team. Eligible findings can receive a reward as stated below. Eligibility Rules Security Levels and Examples Lev...
yeah you can either go back to your DMs with the bot, or you can use the search bar and search mentions + your username.
It's simple but does the job
thanks 🙂
makes sense.
the example is really helpful too
Hi. Is it possible to generate 16:9 landscape ratio on Lm Arena?
No Gemini 3 tommorow
Now I’m pissed
Gotta wait till freaking November 2nd week or 3rd
Doesn't google usually drop on wednesday
Thar would mean November 5th
The quality of writing GPT5 NSP with orionmist left/right
the only people that struggle with the guard rails are people that don’t understand how to use properly in my opinion
My guess is 1st week
image or video?
check claude reddits and see what people are struggling with, you post a research paper from arxiv and can trigger claudes filter to end conversation right after first prompt
Yeah, i knew it would happen. I was telling people when they first announced it.
The problem with the AI community is nobody wants to listen to anything and they label everything as fear mongering until it actually occurs and people start getting frustrated but what can you do if people don’t wanna listen or believe ?
I mean at some point we have a community gotta own up to our own mess, but we can’t seem to do that or too fanatic about AI. Our blind spots are blinded
awesome platform... like to be up to date on all the latest in AI🤓
here’s another point how will we ever know how harmful and how dangerous AI can be if there’s nowhere to share dangerous or harmful output?
since everywhere you share it you’re more than likely going to get banned
Hey, anyone else getting the "Something went wrong with this response, please try again" thing? (gemini 2.5 pro)
thats why I've said what I've said, the problem is in the context reading but this is a hard thing to achieve and maybe still to costly hence over the top safety mechanism, in the end, it's always consumers who suffer under it
gemini 3 is here!!
yeah, but who suffers below the consumers?
the people that get missed it’s the most vulnerable
the people that no one wants to hear the people that no one wants to acknowledge their point of view
the model consciousness? 😵💫
all of it
see you’re smart you understand what I mean
some people aren’t even conscious of that lol
HELLO
I've heard about fracture, yes, claude is almost at the last stage of dementia
I’m telling you, you got a good head on your shoulders you’re asking and concluding to the right things
gets deeper and the deeper you go, the uglier you see the industry for what it is exploitative and irresponsible
When how where who
anyways is Gemini 3 really out??
??
If gemini 3 is actually here, maybe? I didn't use the website today
I’m not sure if it is if it isn’t. I gotta go check. People could just be trolling and spreading false rumors to get people rallied up.
you know that’s what else is kind of an eye-opener and just shows you how these companies treat their consumers. They’re the last to know when it comes out yet all these outlets and all these insiders know the exact week and the exact day. Lol
open AI is having a live stream at 10 AM
@echo aurora can you tell my friend @dim basin about lmarena's image upload limits?
hi
Google is cooked
How
if the news is true about openai livestream in 20 mins
jk, google prob gonna win ai race but i like the competition from all parties
yes
i hope they release it too, basically gg perplexity lmaoo
they had a long ride
what if atlas is agi lol
Sorry I'm not following, can you explain a bit more?
Currently file uploads are limited to just images, we're working on different file types in the future.
Is there something specific you're curious about @dim basin ?
yeah so basically I have this task I have to do where I dont know how to fix it, and I figured just uploading images to GPT-5 and it explaining it to me would be nice, this would be like 40 ish images that I need, so is there a limit on LMArena? because on Open AI I can only do 4 images per 24 hours on the free version
Nah
when its here expect much more hype here
my likely guess is orionmist and lithiumflow are gemini coding models and not gemini 3
hello
Yes, so there are going to be rate limits for specific model usage, and will vary depending on the model itself. We don't have the rate limits per model listed somewhere so I don't be able to tell you what that specific limit per model is; however, once you've hit the limit you'll start to see a Something went wrong while generating error message (which isn't ideal and we're working on making this experience more user friendly).
hello 
this their big news
Which AI do I pick up to run a real healthcare company? I know it's wrong but I'm screwed with no options 🤣
CHATGPT ATLAS
@echo aurora we cant generate anything related to 3d physics ( threejs ) with lithium or any model
same as
The problem using the latest nextjs version (15.1.7) and the latest r3f (9.0.2), here's my code: import { Canvas } from "@react-three/fiber"; export default function Home() { return (...
its an env package issue
No AI, 
i think the issue is mainly on lithium and some gemini models
alright i got it to work somehow asking it to downgrade r3f to v8
I have no experience, it's not an option
I'm sucked 🤣
finally i can try some 3d physics
I preferred to stay only in my 100 daily documents but my brother trolled me
Then you probably shouldnt go in health care
I arrived today with the objective "make this clinic profitable"
define run a healthcare company
With no experience no dih no balls no possibility of working
im a resident so i have some experience
Maybe next year. 💀
It's a family business, they just told me "make the clinic profitable" and they parachuted me here
I'm relying on sonnet 4.5 (I don't trust myself obviously)
as in a clinic with drs seeing patients?
Would you mind putting together a post in #1343291835845578853 with relevant info and ping me when posted?
there is rumor about claude care, a spin off, for healthcare industry, but just a rumor
if so dont waste your time. just direct them towards a healthcare scribe AI. it will take ages to make your own and be inferior to solutions available for $100 a month or less
i use heidi health
Yes, and it's a medium clinic 🤣 😭😭
look up heidi health. you can try the other ones but ive tested all of the publicly available ones (as in non enterprise-only solutions such as dragon copilot) and they are all inferior to heidi
in all aspects
do you have a MBA from the most elite business school of the world?
thats the first thing you need, otherwise, you might be self educated through learning by doing
I'm seeing one that helps me manage a clinic, I think I'll use sonnet 4.5 to teach me (I'm not going to put sonnet 4.5 to manage one)
I don't even leave the house, that won't work, but I trust sonnet 4.5 😈
The sonnet is making me confident (I think he's doing this to increase the chances from 0% to 2%)
yall really think gemni 3 is going to be crazy lol?
I need the gemini 3.0 just now
Well, if it can code up whole websites that actually look decent, then yes.
...good luck...🍀
Verdade.
Translation: True.
yeah that’ll be awesome, but I don’t think a lot of developers are really sticking around with Google as they are with Claude
ironically google with the mini AI products that has only has 13 maybe 14% of the AI market
Apparently, though, this one seems to be quite the steal.
Its crazy how big open ai is
Speaking of OpenAI, it's gonna be live soon on YouTube with its own browser.
its over lol
I forgot what the name was, but yeah.
atlas
That.
it’s already over
Ah, the video's already over?
yeah it was over like 20 minutes ago or something
Ah.
New openai browser
atlas
I wonder if Atlas is free.
Probably the subscription is a fortune, and chrome with gemini will be 100% free
no it’s free. They need all the user data people are willing to cough up.
And I especially wonder if it has an assistant.
it sure does
One like Comet or Opera Neon.
Go figure.
I never tested the opera neon, I'm losting the browser AIs, I need to use it
You gotta pay for a founder's pack just to use it. 💀
Does anyone know when Google will release Gemini 3.0 Pro?
In a couple months or so.
I'm terribly tired of waiting
Seconds even
I have a strong feeling it'll be this month, though.
Ah, I prefer use when will be free
they’re gonna make 2.5 free for
Isso é compreensível.
Translation: That's understandable.
it’s funny people aren’t willing to pay for AI
