\
  The most prestigious law school admissions discussion board in the world.
BackRefresh Options Favorite

Gemini 3 destroyed all other LLMs

...
Crusty Provocative Coldplay Fan
  11/24/25
...
abnormal bonkers theater foreskin
  11/24/25
...
Crusty Provocative Coldplay Fan
  11/24/25
won't Altman and co just focus on subsidies via AI data cent...
insecure parlor cuck
  11/24/25
I was thinking about this too Anybody here have any exper...
Thriller field
  11/24/25
I just use the gemini app on iphone. What does a google phon...
abnormal bonkers theater foreskin
  11/24/25
I think it just lets the Jews/Indians hoover up more of your...
Thriller field
  11/24/25
...
nudist weed whacker
  11/25/25
looks like Opus 4.5 might be a better coding model now. hard...
nudist weed whacker
  11/24/25
...
insecure parlor cuck
  11/24/25
Claude is nigger af ime and the limits are fucking ass outsi...
honey-headed autistic heaven
  11/24/25
GPT 5.1 did a little better on Biglaw Bench than gemini 3. I...
abnormal bonkers theater foreskin
  11/24/25
gemini 3 is retarded sick as fuck, im stealing 9$ mill PE fu...
honey-headed autistic heaven
  11/24/25
...
abnormal bonkers theater foreskin
  11/24/25
bro i just want to be left alone ina room with access to gem...
honey-headed autistic heaven
  11/24/25
...
abnormal bonkers theater foreskin
  11/24/25
translate from nerd to english please
Slimy frozen university trump supporter
  11/25/25
I am skeptical about the value of comparing models on benchm...
kink-friendly senate laser beams
  11/24/25
probably but they just released it today so I haven't had ti...
Fragrant station
  11/24/25
Nano Banana Pro Create an image at 31.7785° N, 35.22...
insecure parlor cuck
  11/24/25
Why is this either impressive or useful
fantasy-prone stimulating boltzmann hall
  11/24/25
the way humans will learn, create, and consume content all l...
insecure parlor cuck
  11/24/25
It’s so much better than other generators.
kink-friendly senate laser beams
  11/25/25
https://x.com/karpathy/status/1992655330002817095?s=46 Th...
kink-friendly senate laser beams
  11/25/25
...
insecure parlor cuck
  11/25/25
gpt 5.1 and grok 4.1 are both better than it. at getting ans...
Fragrant station
  11/24/25
Better at what? Destroyed it at what? I have had it on my ...
Lascivious garnet gas station
  11/24/25
Have you tried suing people with it?
abnormal bonkers theater foreskin
  11/24/25
it just came out like 6 days ago
Fragrant station
  11/24/25
They all miss important details when they review legal docs
Ivory Nursing Home Immigrant
  11/24/25
Not if you train it in notebook lm
honey-headed autistic heaven
  11/25/25
Not if you specifically feed it everything it needs to the e...
outnumbered water buffalo
  11/25/25
...
Out-of-control address hominid
  11/25/25
yeah but feeding things into an llm is much easier than rack...
Fragrant station
  11/25/25
...
Out-of-control address hominid
  11/25/25
cr. this is why I enjoy vibe coding even if doesn’t sa...
nudist weed whacker
  11/25/25
It’s amazing how committed you are to this schick.
outnumbered water buffalo
  11/25/25
...
Crusty Provocative Coldplay Fan
  11/25/25
TPUs can do anything anyone would want to do with neural net...
nudist weed whacker
  11/26/25
Sorry ur too Poor to afford Opus 4.5
Out-of-control address hominid
  11/27/25
Claude opus 4.5 is legit but I think Gemini 3, opus and gpt ...
Fragrant station
  11/27/25
This has to be a joke. Other than synthesizing massive infor...
hyperventilating new version
  11/27/25
It outperforms the other leading models in certain metrics&m...
Fragrant station
  11/27/25
There seems to be a limited amount of benchmark gaming going...
nudist weed whacker
  11/27/25
Yea makes sense since deep mind has been doing all of this s...
Talented Useless Brakes Alpha
  11/27/25
...
insecure parlor cuck
  11/27/25


Poast new message in this thread



Reply Favorite

Date: November 24th, 2025 1:29 PM
Author: Crusty Provocative Coldplay Fan



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49456404)



Reply Favorite

Date: November 24th, 2025 1:34 PM
Author: abnormal bonkers theater foreskin



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49456421)



Reply Favorite

Date: November 24th, 2025 1:59 PM
Author: Crusty Provocative Coldplay Fan



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49456522)



Reply Favorite

Date: November 24th, 2025 2:02 PM
Author: insecure parlor cuck

won't Altman and co just focus on subsidies via AI data center construction and deployment all over the world to help lower the birth rate?

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49456528)



Reply Favorite

Date: November 24th, 2025 4:16 PM
Author: Thriller field

I was thinking about this too

Anybody here have any experience with the jewgle phone?

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457082)



Reply Favorite

Date: November 24th, 2025 5:57 PM
Author: abnormal bonkers theater foreskin

I just use the gemini app on iphone. What does a google phone add?

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457429)



Reply Favorite

Date: November 24th, 2025 6:54 PM
Author: Thriller field

I think it just lets the Jews/Indians hoover up more of your data so you can Do The AI better

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457590)



Reply Favorite

Date: November 25th, 2025 10:03 AM
Author: nudist weed whacker



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49458928)



Reply Favorite

Date: November 24th, 2025 4:13 PM
Author: nudist weed whacker

looks like Opus 4.5 might be a better coding model now. hard to know for sure, but it's close. the rate of advancement here is nuts. pretty decent improvement over Sonnet 4.5, which isn't even 2 months old.

https://www.anthropic.com/news/claude-opus-4-5



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457074)



Reply Favorite

Date: November 24th, 2025 5:14 PM
Author: insecure parlor cuck



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457269)



Reply Favorite

Date: November 24th, 2025 5:58 PM
Author: honey-headed autistic heaven

Claude is nigger af ime and the limits are fucking ass outside of enterpise subscirption

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457433)



Reply Favorite

Date: November 24th, 2025 5:58 PM
Author: abnormal bonkers theater foreskin

GPT 5.1 did a little better on Biglaw Bench than gemini 3. I haven’t done a Shitlaw Bench yet

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457434)



Reply Favorite

Date: November 24th, 2025 6:00 PM
Author: honey-headed autistic heaven

gemini 3 is retarded sick as fuck, im stealing 9$ mill PE funded companies workflow and sites staight from them with Gemini and Framer HTML Chrome extension, this is fucking insane I should be hunted down by these companies cant even imagine what kind of evil shit is being done with this stuff its 180

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457439)



Reply Favorite

Date: November 24th, 2025 6:03 PM
Author: abnormal bonkers theater foreskin



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457455)



Reply Favorite

Date: November 24th, 2025 6:06 PM
Author: honey-headed autistic heaven

bro i just want to be left alone ina room with access to gemini 3 and notebooklm and my last decade of google drive docs

thats it, i got it from here i dont need anyone or anythign else

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457462)



Reply Favorite

Date: November 24th, 2025 8:29 PM
Author: abnormal bonkers theater foreskin



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457921)



Reply Favorite

Date: November 25th, 2025 10:09 AM
Author: Slimy frozen university trump supporter

translate from nerd to english please

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49458946)



Reply Favorite

Date: November 24th, 2025 6:35 PM
Author: kink-friendly senate laser beams

I am skeptical about the value of comparing models on benchmarks where they cluster together at high scores. There is not much room to detect model differences when they are at 90 %

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457525)



Reply Favorite

Date: November 24th, 2025 8:42 PM
Author: Fragrant station

probably but they just released it today so I haven't had time to test it

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457960)



Reply Favorite

Date: November 24th, 2025 6:39 PM
Author: insecure parlor cuck

Nano Banana Pro

Create an image at 31.7785° N, 35.2296° E, April 3, 33 CE, 15:00 hours

Thinking...

Examining the Details

I'm currently focused on the specific coordinates and date provided. My goal is to produce an image reflective of that location and timeframe. I am now working on identifying relevant historical data points.

Pinpointing the Scene

I'm now zeroing in on the specific elements needed for the historical recreation. The latitude and longitude lead me to Jerusalem, and the date focuses me on a well-documented event. I'm building a picture of the scene, considering details like Roman soldiers, a crowd, and the crosses on a hill, all against a fittingly somber backdrop.

Analyzing Visual Depiction

I'm currently reviewing the match between the visual elements and the user's input, specifically focusing on the accurate representation of the specified time, date, and location. I am ensuring that all the parameters are correctly translated into the visual depiction. The focus is on the correct elements of the crucifixion scene itself.

Assessing Scene Correspondence

I've determined that the generated image aligns completely with the user's input. The specified historical setting, time, and date are all evident in the scene. The elements described in the request are correctly reflected in the visual output. The historical context, Roman soldiers, and onlookers are consistent with the prompt.

https://imgur.com/a/ZYhQE7s



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457550)



Reply Favorite

Date: November 24th, 2025 8:33 PM
Author: fantasy-prone stimulating boltzmann hall

Why is this either impressive or useful

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457939)



Reply Favorite

Date: November 24th, 2025 8:47 PM
Author: insecure parlor cuck

the way humans will learn, create, and consume content all look to involve Google moving forward.



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457977)



Reply Favorite

Date: November 25th, 2025 1:15 AM
Author: kink-friendly senate laser beams

It’s so much better than other generators.

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49458420)



Reply Favorite

Date: November 25th, 2025 1:20 AM
Author: kink-friendly senate laser beams

https://x.com/karpathy/status/1992655330002817095?s=46

They must have trained it with Gemini because it can solve complex technical problems in the image generation process.

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49458422)



Reply Favorite

Date: November 25th, 2025 10:08 AM
Author: insecure parlor cuck



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49458944)



Reply Favorite

Date: November 24th, 2025 8:41 PM
Author: Fragrant station

gpt 5.1 and grok 4.1 are both better than it. at getting answers wrong.

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457957)



Reply Favorite

Date: November 24th, 2025 8:45 PM
Author: Lascivious garnet gas station

Better at what? Destroyed it at what? I have had it on my phone for months and after some initial fun making pictures I basically stopped using it.

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457969)



Reply Favorite

Date: November 24th, 2025 8:46 PM
Author: abnormal bonkers theater foreskin

Have you tried suing people with it?

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457972)



Reply Favorite

Date: November 24th, 2025 8:47 PM
Author: Fragrant station

it just came out like 6 days ago

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457973)



Reply Favorite

Date: November 24th, 2025 8:46 PM
Author: Ivory Nursing Home Immigrant

They all miss important details when they review legal docs

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49457971)



Reply Favorite

Date: November 25th, 2025 1:09 AM
Author: honey-headed autistic heaven

Not if you train it in notebook lm

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49458417)



Reply Favorite

Date: November 25th, 2025 4:01 AM
Author: outnumbered water buffalo

Not if you specifically feed it everything it needs to the extent you could have just done it all yourself tp

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49458504)



Reply Favorite

Date: November 25th, 2025 4:04 AM
Author: Out-of-control address hominid



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49458509)



Reply Favorite

Date: November 25th, 2025 4:05 AM
Author: Fragrant station

yeah but feeding things into an llm is much easier than racking your brain doing all kinds of stepwise work yourself and it feels more rewarding because its like you are cheating. well to me anyway cheating feels more rewarding than actually doing something.

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49458511)



Reply Favorite

Date: November 25th, 2025 4:06 AM
Author: Out-of-control address hominid



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49458513)



Reply Favorite

Date: November 25th, 2025 10:02 AM
Author: nudist weed whacker

cr. this is why I enjoy vibe coding even if doesn’t save time. I would rather just keep playing the code lottery and ask it to regenerate than have to think

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49458925)



Reply Favorite

Date: November 25th, 2025 4:00 AM
Author: outnumbered water buffalo

It’s amazing how committed you are to this schick.

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49458502)



Reply Favorite

Date: November 25th, 2025 3:45 PM
Author: Crusty Provocative Coldplay Fan



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49460011)



Reply Favorite

Date: November 26th, 2025 12:24 AM
Author: nudist weed whacker

TPUs can do anything anyone would want to do with neural networks.

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49461564)



Reply Favorite

Date: November 27th, 2025 2:57 AM
Author: Out-of-control address hominid

Sorry ur too Poor to afford Opus 4.5

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49465135)



Reply Favorite

Date: November 27th, 2025 4:20 AM
Author: Fragrant station

Claude opus 4.5 is legit but I think Gemini 3, opus and gpt 5.1 are all good and different things. I still like GPT best overall though.

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49465164)



Reply Favorite

Date: November 27th, 2025 3:08 AM
Author: hyperventilating new version

This has to be a joke. Other than synthesizing massive information contexts for simplistic Google search style prompting, or analyzing a video on YouTube in detail, it is still complete trash compared to GPT and now Claude

Still no projects or OEM prompt optimization tool. Looks like they removed agentic prompting

Today it could not review a 4-page PDF in order after 4 tries (kept saying something on page 1 was on page 3).

Tempted to try Ultra but honestly why bother

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49465146)



Reply Favorite

Date: November 27th, 2025 7:53 AM
Author: Fragrant station

It outperforms the other leading models in certain metrics—particularly ones AI classically have a hard time with but this doesn’t mean it’s actual performance is better. Labs are starting to use internal proprietary metrics and a lot of the known public ones are now saturated or gameable

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49465233)



Reply Favorite

Date: November 27th, 2025 11:52 AM
Author: nudist weed whacker

There seems to be a limited amount of benchmark gaming going on. Private benchmarks or new benchmarks released after the models come out usually produce a rank ordering similar to other benchmarks. The labs have an incentive to be straight about benchmarks because people quickly figure out what models actually work and they won’t trust companies that consistently game them.

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49465528)



Reply Favorite

Date: November 27th, 2025 9:06 AM
Author: Talented Useless Brakes Alpha

Yea makes sense since deep mind has been doing all of this stuff we're seeing publicly since 2010

(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49465287)



Reply Favorite

Date: November 27th, 2025 9:41 AM
Author: insecure parlor cuck



(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2most#49465328)