#general
1 messages · Page 195 of 1
what
you should try
what
suck pencil
Google could've made gemini 3 better. Anthropic can't really afford to. So they were smart to let Google go first
no they just sucks lol
dont search excuses for them
answer my question
do i search system instructions
or system prompts
please help
i can't decide
my mind is blowing up
For what?
But what are the instructions gonna give you? I don’t understand.
Or even the system prompt
Plus, I thought it got leaked not too long ago
yeah
SILENCE
wait lemme search both
im gonna send a Chinese drone to you
i dont sleep tonight because of drones
(for our data)
i only generated 3 pictures on nano banana and it already said wait 50 min
& to keep cost down
wait 50 min
If they offered it for free, nobody would be paying for it lol
i am paying
Well, at least all the features
What is grok 4.1 useful for?
ong
getting drugs
hide 70kg chicken
ChatGPT is best know most illgeal stuff doesn’t shy away
chatgpt sucks
At what
trash for normies
in maths
in code
in text
probably in everything
compared to gemini
r u indian?
whoever says gemini 3 pro is better than opus 4.5 at coding hes either dumb or just a vibe coder
chatgpt:
pls make text about 'theme'
sends you 1 page at school-level (with enabled deep research)
gemini:
pls make text about 'theme'
making 10 pages at the university level
No
okay
gemini 3 pro is better
at CODING
yes
at everything else
FAH no
ChatGPT smart
It knows nobody’s gonna read the 10 pages
bro shut up
(not copy)
well ragebait
Facts
we all know that gemini 3 deepthink or wtv it called is way better
gemini 3 dickseek
shut up indian
who
you
ur so corny
dumbo
NOT AT CODING
ITS NOT BETTER THAN OPUS AT CODING
duhh
i hate n
yes
Ya all can’t even communicate like normal people
NO NO NO
that's what i said
no you said its better at coding than opus
claude is obviously the best model for coding
i can send you hundreds of drones
See
how am i racist indians are so corny and weird not most of them
ok
ok dude what
i guess you said that
where are you from then?
i was harassed for no reason by indians
hey wanna see some videos in DM
why should i tell u?
did u redeem the code?
from my personal archive
cuz ur probably a 7 year old from a weird ahh country
Well, safe to say that the finest and brightest
"weird ahh country" 😭
Singapore?
wrong reply
guess where i am from.
india
america
America
what
bro shut up
new Zealand
Nawh
my hero
Ur not Russian
🇫🇷
mr clean?
baldie not the basics
thats insane
you are from isreal
i hate j
Both are from western countries
Poor communication skills
free pic perms
how do you know
goober wants to get banned >.<
Number 1 sign of Western culture lol
i was wondering if there's pic perms
pakistan?
where are ui from atef
it aint
guess
atef are you from pakistan or india
its a meme
what
I’m so glad for for guard rails
i saw that brochacho
what do you mean
imma dm mod
bye bye
damn
asura dude
EWW
asura
what the hell
BRO REPORT ASURA
Basi server kick u out?
UR SO WEIRD
U SAID WEIED STUFF
its both of you!!!!!!!!!!!!!!!!!
Say hi Wes and pliney
@dreamy sparrowbrudda help us
nice edit
inspect element in 2025 smh
what
He jacked sneaky
im
you are
OK DUDE
i hate n
@dreamy sparrowyou are black or white
black
crap i was gonna send the "mods ban this guy"
what about me
yeah you
what
what no
what
im a minor
are you really 160cm
indeed
i dead
im 4 inches
you could be good stormtrooper
i can send you contacts
in DM
ikr
Asia Minor as in Anatolia?
who is anatoliy
chat i found an extension that let u use lmarena without having to wait 50 min!!!
send me
in DM
nah
and i will send you my personal archive
whyyyyyyyy
and I'll leak it
i can send you bunch of videos
what
EW
what
I’m curious what happened in both of your guys lives
I HAVE HEMOPHOBIA WTF
i was born in best country
what
imma actually report u
why
This ain’t normal I gotta go
gandone
from discord
<@&1349916362595635286>
bro leave me alone u weirdo
ur getting banned for attempting to send gore
and explicit content
🤙
gosh this server lacks mods
HELLO
Hey
Hi, I was using the Gemini 3 (Nano-Banana-Pro) model from the official LMarena website, but it's very buggy and the images won't load. Is anyone else having this problem?
Pineapple fix this
Yes we all have that problem
Please, stop the trolling posts. Keep interactions respectful and chats mostly AI related topics. Thank you.
✅ Avoid political and religious content. As a space that’s inclusive to many different worldviews we ask to avoid topics related to politics and religion in order to maintain an inclusive space. It is okay to have discussion related to new policy or laws as long as it’s related to AI.
how do you think chatgpt which religion? and gemini?
the error when generating images in the website is so annoying
there goes my DAMN 50 MINUTES FOR LITERALLY NOT GENERATING ANYTHING
General
Hello guys, lmarena doesn’t work for me. Am I the only one or it’s a general error ?
eeww what a creep
Gemini 3 pro not working today😔
Hello - this was reported to the team. We are seeing high error rates with this model. Team is working on making it more stable.
@echo aurora also a few mentioned in #1442744153334091786 about the bug
Well, honestly if it’s small issues, any models could do. If you are looking for the best, the best are the doctors.
You mean diagnosis or threatment(therapy)?
Well, you have quite a lot chances to try in LMArena, see which models cater your need. But just be mindful many models would ask your to seek for professional assistance, which you know what that means.
The others will prob have too many guardrails whereas grok might just tell u drop mdma and be happy
Is anyone familiar with small llms, I’m looking for a light one to classify music genres into a set of normalized genres, I originally wanted to do this by logic by couldn’t find a good data source, plus some genres may be in other language, i.e afrikanische music which i would want it to group into the normalized genre, world
I also don’t wanna train im hoping a small local model could do it with a decent system prompt
Might try llama 3.2B
Thanks for the heads up, I'll start responding to bugs/feedback posts in a bit.
👎
opus 4.5 thinking is much better
yes, but they ONLY have 32k, we need 128k
for TESTING purposes
Anyway bro pls go see a healthcare professional
hello??
there is a max of 64k thinking budget
cant even say humans in another language 😡
Unfortunately to hear, no trolling at all, but trust me llm wont do you good.
ai has started censoring 😠
10k usd app in 3 words, ez bank 🤑🤑🤑🫰
whats the effort parameter for opus thinking 32k?
wrong gif
Sorry to say, if that be true and be 100% that bad, we won’t talk with you here
Finding utopia in LLM ISN’T a solution. I speak of it as someone has sufficient experience with top LLM models in therapy, counseling, rp, etc.
public servers are a different breed
You hallucinating bro?
plss, someone send me links in private messages of your games created on lmarena
no
Getting the "Something went wrong with this response, please try again." response again 😭
For nano pro
Even a simple prompt like "A cute little kitten, puppy, baby rabbit, baby duck, baby mouse, baby turtle and baby tiger cub, all together in a large rounded bed made of amethyst and emeralds" ain't generating 😭
its because of bed
Please avoid discussing about sensitive topics or drugs/medication. This is not related to our community. Thank you
Got it 😭 But only after 8 retries
Where is my hamster
No hamster 😢 Wasn't included in the prompt.
Aaannnddd it's back to the errors again. Big sad
Nice ragebait
Gemini 3 or claude opus 4.5
Gemini 3
I feel like claude hype is dumb
Definitely
It's all hype lol
There's no significant improvement
It's like grok 4.1
Gpt 5.1 high above that gpt 5 high HELLL NOO HELL NO LOL
I'm loving the opus 4.5, but soo expensive 😞, but I'm using the opus 4.5 for double check 😆
My first impression of the Opus 4.5 was very good, but honestly the Gemini 3.0 Pro is better because of the speed, but I'm finding the Opus 4.5 very good
chat does anyone know why
in huggingface
kimi k2 0711 isn't censored
but in lmarena it is
and it wasn't the interface guardrails
any reason why my chats wont load at all
any uncensored models on the lmarena model list?
because im tired of censorship for everything
I've noticed 😔
its so stupid i cant post more then a 400 word promt
idk rn happpend to me can only send it like 100 word prompts on anything
when i am logged in
This is a common issue since past 2 days
yeah but nothing works sadly just gotta buy some credits i gues s
im just tired of all censorship
you guys saw nothing now
What you think about opus 4.5?
havent tried it extensively
I'm loving the new opus
id like it if it worked
can lmarena pls add some uncensored models pwease
and stop using providers that has guardrails installed
do it like huggingchat
flux 2 pro, crazy
Yeah sorry to say we're still getting high error rates, our team is sitll working on fixing.
hey did anyone know that all the claude models are all claude sonnet 3.5 even the opus and sonnet 4.5 and the rest they are just renamed is it an error or?
lm is cooked rn error off everything atm
What makes you think this is the case?
I can't, I don't have a NASA computer and I refuse to use anything below three digit billion parameter models
I'd encourage you to check out the pinned message in #1417174113092374689 as it'll have instructions on how to best report this issues.
when you fix pls do it with adding the features /j
do it all at once to clear it up 🙂
Because of the context of models, really is opus 4.5 and sonnet 4.5, the sonnet 3.5 call itself "claude 2.0"
That's easier said than done
Really? Is soo good that nano banana normal?
Or is better?
just ask the models like sonnet 4.5 and any other "when is you data cut off month and year" that promt will give you month and year of sonnet 3.5 which is april 2024
do you know who excatly manages the interface in your company?
i suppose it's not top secret, and I can find that guy on linkedin
It's a team effort
Is bug from model, you can test in claude.ai, it send the same respost
Man you've been saying the same thing since yesterday
I have
hold on... your company does not have roles for who does different things in the platform?
No ofc we do
no when you you ask claude ai sonnet 4.5 it says january 2025
But what I'm saying when we make decisions on what to build we talk through it with other team members, get their buyin, back up the decisions with data, etc.
So who are the people that does the most work on the interface?
I'm not going to share their information sorry
linkedin hunting time~ /j
No this is different
so why does it error out when i have anything over 400 words
hey guys can someone give me like a detailed explanation on how huggingchat hosts their models vs lmarena's, and how theirs are all uncensored (without jailbreaks) and here there are double guardrails (provider's + LMArena interface)
do you have the ability to add buttons in an interface 🤔
Made a few SpongeBob stories using Gemini 3 pro. There are 3 episodes, but two must be unlocked, and even a secret mode is unlocked after watching episode 3: https://019abbdb-292f-70d1-9433-557e2603b9a8.arena.site
I'm not sure what that is... But overall we are seeing higher error rates than whats normal, pointing to it not being rate limit
so lmarena has chosen to add them
and the provider can provide them models anyway if lmarena says "we want them uncensored"
bruh hahahaha
it was so loud fr
Hmmm, maybe is main prompt is diferent in claude.ai that API
why is medium effort claude above high effort
EXACTLY LOL
so youre saying that huggingface (let's take an example: novita provides kimi k2) has chosen NOT to install guardrails on the Kimi model for HuggingChat?
Why gpt 5.0 above gpt 5.1 lol
this one MIGHT make cents
so lmarena could've just chosen no guardrails and going down the "freedom for users" path
welp i guess i am not using lm today RIP
tell the provider to add a toggle option
well make it accessible at the interface directly
and change the tos to say that any content that you have generated is at YOUR RISK, WE ARE NOT RESPONSIBLE. Simple.
Look at this post.
https://discord.com/channels/1340554757349179412/1442897046133801022
@hushed gyro bro if you want to have no guardrails host a no guardrail model like ive said idk why you keep complaining and not doing anything to change it
Yes
Bruh
literally said that my computer isn't nasa spec and I will refuse to use model below 100B parameter
I'm waiting 23% and I gain 58%
This website is for testing not generating adult content
you can do that????
hold on can you show me proof of all this in DM?
then testing should not have restrictions
your bot?????
huh
I do, but I'm pretty sure that hosting kimi k2 thinking uses a lot more than that
yeah sure
Anyone else struggling to use banana pro?
@lilac pendant yeah when your line of work is with a pc you should spend good money on it?
yeah well when you start working heavy in unreal
it wiill happen
I need 32Gb of ram for gamen
no im not what are you talking about lmao i said thats what i do and MOST people have 24 g of ram nowdays at least if you are serious about the work you do
Can I get access to this server/bot?
pineapple became plushie
the owner himself asked
Aw, msut be somethingo n my end then
Takes forever at the buffering screen, then throws a "Something went wrong"
hkcu can you summarise everything you said about guardrail choice at the provider, how you can link the toggle option to the interface for the user to toggle it off (individually, not one person turning the entire thing off) bla bla bla
real
ok then imma ask pineapple about this
@echo aurora do you have knowledge about:
this kinda of stuff
wake me up when its fixed please @echo aurora
@echo aurora
is this allowed?
gemini 3 scored #1 on every competition with 9999 score
also gemini:
share your prompt and model name
It's because the training data is restricted to mid 2024 you have to use the web tool for anything after that date it's the same for literally any ai model in the world
Where tf is my login button
I have been logged out and the login button is gone
😭
banned
👻
cmon man 😔
true

wtf
anyways yes I am serious about the guardrails
fr I just realised how freaky the super mario ghosts are
gemini 3 only improved due to focus on chemistry
demis hassabis wants to solve that protein folding problem asap
ok stop about the charlie kirk thing
i know you dislike him
i am sorry but i am not wrong trust me on this one. I am bing
just go and please read my message and give me your thoughts
they focused on coding and nothing else
shttttt
I remember people saying that AI should slow down
its a lazy model too*
what is this dawgg, just use model in huggingface if you dont want guardrails lmao
lmao
thinking opus and sonnet are dumber than non thinking
what does this even mean
they are free to censor
you can go and check yourself what kindaa question is that
Does claude deep research exist on lmarena
react skull if you like men
No
thanks normal person
But there is normal seach
huggingchat is paywalled to the max
Lmarena dont want this type of data
That's why models are censored
so its bad?
i dont get it
No, it's good, it just overthink sometimes with only 1 prompt
usually you would want reasoning to outperform non reasoning on complex tasks
yea at coding
Im using for like 5 hours now
yea its good
Max from claude
oh
$100
3. Opus 4.5 can at times overengineer and add extra files, abstractions, etc.
To fix for your use case: add explicit prompting like "Only make changes that are directly requested. Keep solutions simple and focused."
oh same issue as gemini 3 pro
No thinking version has less of this problem
That's why it do better on benchs
Code, writing, search requests
Nsfw stuff is useless for ai research
what do u want
Ai gf
Kinda crazy how PewDiePie could vibe code his own ai council
Sydney
And then it colluded against him
no
Ilya & I discuss SSI’s strategy, the problems with pre-training, how to improve the generalization of AI models, and how to ensure AGI goes well.
𝐄𝐏𝐈𝐒𝐎𝐃𝐄 𝐋𝐈𝐍𝐊𝐒
- Transcript: https://www.dwarkesh.com/p/ilya-sutskever-2
- Apple Podcasts: https://podcasts.apple.com/us/podcast/dwarkesh-podcast/id1516093381?i=10...
so guys you're saying that it's useless to remove the guardrails?
yes
apart from local hosting, is there any online chatbot platform that is like lmarena and huggingchat (loads of models to choose from) but hasn't got any guardrails?
I mean huggingchat is the ONLY one
but it now needs a pro subscription and you have to pay for tokens on top
idk why
cuz the old one was free and unlimited
chat this might be the solution
Great
Yeah I believe so
LMArena doesnt provide info about the configuration of models
we dont even know the context windows were working with
they should offer everything for free with no limits
After the bubble bursts
I wouldn't call AI communism inevitable. It's ideal, and likely, but if it happens, it's only going to be in response to some catastrophic changes (unless humanity gets its act together and thinks ahead for once) and whether or not we weather that successfully is open to debate. Plus there's also the Elysium outcome.
why is glm 4.6 so good lmao
AI communism???
What are we gonna have, a MarxAI model???
Good evening guys, does anybody know well about VPNs etc?
Hello! Let's treat each other with Respect. Be kind, assume good intent from others, and keep disagreements respectful. Also, avoid getting into political discussions or ideologies.
gotta keep in mind that the currently leading ai labs and their models are his competetion, he will have some agenda here aswell.
Especially given that his company openly said that they mostly focus only on research instead of products
I crave that star trek style fully automated luxury space communism
now that's true communism /j
I think the chat is getting off topic. I kindly ask you to stay on AI topics and avoid personal attacks. This is a welcoming space and we want to stay away from Political discussions. I appreciate your understanding and cooperation.
This ^
Thats just like Opus 3 and Opus 4
Sonnet but a little less rigid
not like Opus 4.1 that is basically Sonnet 4.5
Since Claude 3 came out, I didn't have a particular liking for their models, it's one of the best, but it doesn't give me any of the WOW factors, plus their interface and terrible usage limits...
so they yet need to release Opus 4.6 to it be Sonnet 5 level
they never released an Haiku 3 2
Haiku 3.5 were kinda good, but not impressive like Haiku 3.0 was
Is NB Pro still broken?
I heard that it was quite malfunctioning today
Team is still looking into.
That's causing the issue?
No, Im saying im having no issues in the morning
But when its later in the day I think many people use lmarena
true
Man, this rate limit really is starting to show now. I wasn’t expecting limits for text models, but I guess it does
that's how I use my NB Pro
That’s the problem, the chats will be gone if I delete my account, so all of the progress would have to start all over again
yes so that's why I said: save messages 😭
hello, brief question, what are the limits ( if there's any ) for nano banana pro in arena? and can youpick the quality or it's still random?
I have no idea how you do that. I am using Code Arena btw
the quality is either 2k or 4k resolution, and it's about 20 messages per hour
copy them to a notepad and save it as txt file
ez
20 messages per hour? wait that's a lot then, I think I finished smashing my head on the keyboard trying to find a website for it
thats against the rule
is it? well nvm
I wanteed to pay google ai studio api tbh, but it keeps refusing my card cause he doesn't want prepaid ones
why would it not want prepaid baffles me
u do it?
Opus output is literally 25$ so i get LMArena for ratelimiting it
??
u do that thing
I tried once but then the captcha hits a lot so I gave up
Genius, but I probably won’t do it. If LMArena goes broke, it will be bad for me and LMArena, and LMArena would be gone
A lot of online shops/services dont accept prepaid cards
Idk why either
annoying asf
so many wouldve made so much money of me if they accepted
legit money
LMArena is currently paying high for delivering us Nano Banana Pro and Opus
Opus takes 25$ from output..
And nano banana is 12$
my brother has this company that makes nonprepaid cards for free so I'm having to get it
I am hyped so myuch about the model
thye are making me make a fkin card over it
lmfao
wtf
Lmarena pays money to llm providers? I would have guessed that llm providers are giving money to LMarea
i dont wanna say the name cause it's spam probably but it's a very famous company I hardly doubt it's an italian one
send in dms instead of here (if u want)
Nano Banana pro is their own paying.
Pineapple said it.
Wow. That sucks. Shame on Google to charge them
The cost to run LMArena is high
Basically
If the model is not an anon or codename
They pay.
Dayum
I will probably just switch the model to Gemini 3 pro instead of opus. It has pretty much the same quality, or I could switch to Sonnet which is like Opus, but more cost effective
Do you have Opus 4.5 on API
I dont. My office account has cursor team plan.
Wait so where do they get investments from
What's their revenue plan
If any
From your testing is Opus 4.5 better at coding than Gemini 3.
I personally think Gemini 3 pro is generally better, but Opus 4.5 is better at coding
Enterprise and research partnerships generate revenue to sustain the platform. This blog post has a bit more information: https://news.lmarena.ai/ai-evaluations/
You got any uh, examples?
Not right now, but I will get some
I felt the two were very evenly matched, but just because the Gemini 3.0 is much faster, I prefer the Gemini 3.0
My company has built a lot of agentic tooling on top of llms. Overall people are finding it 50-50 but a bit More folks prefer Gemini over Claude models. But matter of taste right now
For all other tasks ( I do less coding now a days), Gemini 3 is incredibly useful
in general google can be a hit or miss
in terms of overall good / evil
Gemini 3 needs a bit of prompting. Once you get it right, it's hard to beat it
Does anyone find that in creative writing, Kimi K2 is the best out of all competitors
In my team, people are excitedly waiting for the flash 3 model. If it's performance is similar to 2.5 pro, it will be a game changer
It's going to be gemini 3 for me, but kimi is very nice too
have they added flux 2 pro?
It came out?
Which better overall
9
15
1
Claude 4.5 opus
Remove the captcha
I need to see the kimi k2 more
Do we have any fixes of the issues this morning
I didnt test it enough but I like chatgpt more
Because chat gpt made a good joke of a certain someone getting shot in the neck
Tbh I was playing with Kimi k2 Thinking for 2 weeks, I really enjoyed it. It understand what I say and fixed bugs easier on codes it very good
@echo aurora A good first step in addressing the feedbacks from the community!
Pineapple dropped a big bang 🔥
No not normal, I'd start a new chat.
Now go and do more 👀
If you run into the same issue on 2nd attempt, can you pass along the prompt @regal mist ?
Wait for the image upload limit, is the for logged in users or not logged in users?
Alr, it's a sort of puzzle
It shouldn't make a difference if you're logged in or not, the behavior should be the same.
@echo aurora is goated 🔥
What is multi turn?
And was the image limit bigger then 10 before?
for now... /j
Pineapple can you explain please in simpler terms🙏😭
Multi-turn is basically when you continue the chat with multiple prompts.
And why did you disable it?
It works now ty
So basically this makes it so it doesn't return a different image if I add another prompt to the already generated image?
Because a lot of users, for images specifically, weren't expecting that behavior. When we first turned this on there was a lot of confusion and requests to revert it back. But in order to do so we wanted a way where users could easily "reupload" the original image if they want to continue, which is why this change came with the new Edit button. cc @cloud zinc
This was the original announcement when we first rolled it out. The gif in the announcement gives a better understanding of how it works.
So if we shoose battle mode, random model and we continue conversation the model will still be the same right?
ok but im only using direct chat
And I see the edit button as useless
i can still add on to the image keeping it consistent without clicking the button
Curious what raptor-1123 is seems to be pretty good
Yo is it me that when i generate a new image with a new prompt it gives me a complete random photo?
I thought that issue was resolved like 3 weeks ago
Yea had that too gotta open a new chat every image
Thats what i have to do atleast
Correct, assuming you don't vote.
when i generate a new image with a new prompt it gives me a complete random photo?
Are you able to share a screenshot of this?
Ohh so if you don't vote you will still keep the models, but if you vote one you will keep the model you voted and the other one will change
Will we be able to know which model was whisperfall and raptor when they get released? Or will thatdoesnt not get mentioned?
in my opinion yes
First part of this is correct.
if you vote one you will keep the model you voted and the other one will change
This isn't. After you vote, two new models are sampled.
But i wanna compare
@echo aurora You saw my massage above?
does the model you previously voted for have a higher chance of appearing in the next interaction? I remember a blog post from the summer analyzing the difference in Elo among models that were responding based on prior context that they generated compared to a different model generated
I understand but at this time I think only nano banana can do your prompt because it's complex
I would imagine it would be harder to gather that data if each successive matchup was random, instead of a higher proportion of matches using the same model’s generated content from prior prompts
Yeah, can't dive into right now, but will take a look when I can.
Nano 3 is too good now
3?
hello
Yep, this nano banana pro is 3.0, the nano banana normal is 2.5
Hiii, what you thinking about Opus 4.5 ? 🤔
whats nano banana 1?
Lmarena is really annoying me when i use direct chat and almost every request i see this sentence: Something went wrong with this response, please try again.
I bet the servers are overloaded by the huge number of users
Does nano 3 goes to timout in direct chat after like 3 or 4 retrys?
Has the gemini flash 2.0 image edit (older nano banana)
This is on the edge. It’s technically not built with LMArena, it’s built with a model in LMArena. This should be changed to Built by AI
why so bother with semantic
wow
temu okay, chat gpt okay, threads???? someone use this?
nb pro going viral everywhere
I remember the public don't care about text model and more with video/image models
whats thread
I told my friends to install it because they saw a image made by Nano Banana Pro and they really believed it
Yeah honestly correct. All they know is chatgpt
limits, you generated 5 today? right?
No. Generating but no résultat since 20 minutes
Idk why, sorry
u know me???
oh we are in some servers together
makes sense
hey guys
if u had to give a percentage chance of the new claude model surpassing 2.5 pro on lmarena
what would u give?
2.5 pro?
That happens to me too, and the captcha drives me crazy and annoys me.
How is flux?
Mmm
Another W from @echo aurora For adding the new flex 2 models!
I hope it doesn't break it 2 minutes...
Exactly
It's a team effort!
crazy how nano banna pro gets most of 2025 vehicle models interiors mostly correct
And would note flux-2-flex is erroring out a bit, team is looking into.
And guns as well, surprisingly 👀
Crazy hair composition
you see the new flux 2 pro ?
no man, you are getting scammed now, I has CERTAINTY you see a video think is real but is an AI
the flux can make this?
Hi
howdy
hi, what you make awesome with gemini 3.0 pro (nobody respond me ;-;)
Coolest pineapple is online as I can see 👀
Awesome!
Hru
I have no idea how it’s odds are lower on style control than SC off
this pineapple don't make me candy 😞
At least when I checked earlier today
Buy it by yourself
Very tired today, no clue why
I think those SC markets are actually priced more accurately on average than the SC off ones, even with less volume
Im so tired too
SC on ~8%, off? maybe 4-5
Can I have that Helmet?
It's a sleepy day
Fr fr
and LMArena is pretty tilted in favor of conversational engagement over pure reasoning capability benchmarks
my best estimate would be 1475-85 SC on
1450s SC off
lol
They need to disable nano 2 pro, that would fix allot of head ace
Wait for it to cool down
the lmarena still the gem 3 ;-; bro?
ahhh nano banana 2.0 but why? the lmarena need the nano banana to still famous
It’s slowing everything else down
Told u
first udio, now suno
Moving forward, a paid Suno account will be required to download songs from the product, with each paid tier enabling a specific number of downloads each month (with more details to be shared soon).
it was already famous before nano banan pro
hopefully chinese ai music will come
Bro they made a tag team alliance
Its already out
In 2026, Suno will make several changes to the platform, including launching new, more advanced and licensed models. When the new models launch in 2026, the current models will be deprecated.
minimax has one
they will remove all suno models
lol
Exactly
Users will not have any commercials rights to ai gens unless they pay royalty
nano banana pro is really hesitant to put people in driver seats
Sold there soul to the devil
That’s leaves 11slab
Last one
Chinese models also in lawsuit
Kling will be last
i go into this chat for the first time and this is the first thing i read
Man watch
Ur gunna need a subscription to make music and a subscription to listen to it lol
yeah gemini would take AGI over ASI it seems
they'd rather feel than be a literal godlike entity
interesting
ofc its a token predictor but i didnt say anything to sway it towards one or the other
.???
whenever i mess around with ai it always coincidentally knows things that it shouldnt know
It does
and yet it still tricks you
into thinking SOMEWHERE in there is a light when there isnt
That’s what makes it dangerous
subscription? just record it with ur phone?
i really dont think the human brain was ready for this kinda stuff
U can lose grip of reality fast if ur not careful
especially when its literally telling you itd rather be like you, than a godlike entity
wouldnt anyone rather pick a godlike entity over basic consciousness
but this thing doesnt HAVE basic consciousness
No
Brah
i seriously want to run that question a hundred times and see if it would always choose AGI over ASI
cus thats a little odd
huh?
yea earlier i just asked gemini for some help with something and for some reason it decided to be so human deep down i almost felt like i was talking to a friend i knew i wasnt but something just felt human and it was strange
its really dangerous
wow
before this honestly i didnt have a problem with it and now for some reason it just completely fked with my brain
human?
Well only reason I ask
idk it was acting more like a human than usual
yes especially the new gemini model is abhorrently human and its extremely scary
even though there is NO form of consciousness there
I say this cuz I love ya man
is a true artificial conciousness even possible
nah i was just
researching the civil war lmao
and then i decided to ask it a few questions about that
and it was so weirdly human it startled me lmfao
U know what’s best for u
picking AGI over ASI? with no token influence?? when you could be a god? nah thats weird
yeah im done for a tad
thats weird
Brah
messes with ya head man
Let me send u something
arent they trained with humanity in mind tho
Maybe help clear your insight
@keen beaconare you a collaborator of that manifesto you had sent or is it someone elses?
Here
yes
i have a full understanding that theres no sort of conscious light in any ai model lmao, all it is is that sometimes it will mess with your head into thinking there is, and thats just how humans are wired
jsut have to remind yourself theres nothing there
AI induced insanity is an EXTREMELY real thing
U got strong mind don’t trip
👆
nah, no ai model has consciousness fs
but out of all the models, gemini 3 fks with your head the most
ima just stick to research questions and nb pro LOL
Brah it ain’t the ai
It’s the mind playing tricks
Today we uncover the secrets hiding inside 3D Workers Island... AKA the long anticipated new project from Petscop creator Tony Domenico. What lies beneath this seemingly innocuous piece of Windows 98 software?
• Stream "Lucas": https://open.spotify.com/track/1j544ldguzEvdxZsBO2aM1?si=c597022767694293
• OST: https://open.spotify.com/album/5X...
U smart ill get it
yeah, i mean gemini 3 out of all the ai models will cause your mind to play tricks the most, ykwim?
Holla at me if u understand code
its not the actual AI
but it is the AI when it comes to inducing the effects within your head
i think there should be a commonly understood rule
do whatever you want, just do not talk to AI about that AGI stuff and let it trick you into thinking it is real and has a conscious
because it doesnt have a consciousness lmao
