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

With 5.2 GPT seems like it's on top again

Have used the API alongside Claude Opus 4.5 all day every da...
pea-brained cerebral locale newt
  12/16/25
5.2 Pro saved my jobs
irradiated organic girlfriend
  12/16/25
Gemini has been killing it for shitlaw briefs since the last...
khaki lodge
  12/16/25
I'm a Geminimo now.
white exciting corner
  12/16/25
It's 180 because there's basically no lock in effect. Just ...
puce theater
  12/16/25
Exactly especially with the APIs I just switch off between t...
pea-brained cerebral locale newt
  12/16/25
it's like dialing dominoes and pizza hut on a three-way call...
irradiated organic girlfriend
  12/16/25
i'd try gemini, i thought chatgpt was giving the softest, mo...
appetizing stain theatre
  12/16/25
its too safety coded right now for sure. but I'm mostly usin...
pea-brained cerebral locale newt
  12/16/25
totally fair. i'm not using it for machine learning and math...
appetizing stain theatre
  12/16/25
The only reason I use 5.2 is for validating answers from my ...
irradiated organic girlfriend
  12/16/25
It's so ass
Orange doobsian antidepressant drug jap
  12/16/25
I suppose it depends on what you are using it for and also t...
pea-brained cerebral locale newt
  12/16/25
Coding a discord replacement in Assembly language
Orange doobsian antidepressant drug jap
  12/16/25
lol jesus fuck that
pea-brained cerebral locale newt
  12/16/25
Why?
soul-stirring harsh toilet seat
  12/16/25
Performant
Orange doobsian antidepressant drug jap
  12/16/25
What are you nerds even using this Chinese scam garbage for?
soul-stirring harsh toilet seat
  12/16/25
I'm jewish
pea-brained cerebral locale newt
  12/16/25
Python can do a lot and with AI you basically have an on dem...
puce theater
  12/16/25
It gave me an actually good legal memo yesterday (I.e., it l...
low-t mewling stage faggotry
  12/16/25
One of the things that I like about it is it is much more ep...
pea-brained cerebral locale newt
  12/18/25


Poast new message in this thread



Reply Favorite

Date: December 16th, 2025 6:39 PM
Author: pea-brained cerebral locale newt

Have used the API alongside Claude Opus 4.5 all day every day since it came out and they are pretty close but seems obvious to me GPT 5.2 outperforms it

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515064)



Reply Favorite

Date: December 16th, 2025 6:41 PM
Author: irradiated organic girlfriend

5.2 Pro saved my jobs

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515074)



Reply Favorite

Date: December 16th, 2025 6:44 PM
Author: khaki lodge

Gemini has been killing it for shitlaw briefs since the last update

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515085)



Reply Favorite

Date: December 16th, 2025 6:55 PM
Author: white exciting corner

I'm a Geminimo now.

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515110)



Reply Favorite

Date: December 16th, 2025 6:55 PM
Author: puce theater

It's 180 because there's basically no lock in effect. Just use whatever bot is best at the moment.

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515111)



Reply Favorite

Date: December 16th, 2025 7:06 PM
Author: pea-brained cerebral locale newt

Exactly especially with the APIs I just switch off between them and put them in conversation with each other and myself and my outputs and data.

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515134)



Reply Favorite

Date: December 16th, 2025 7:09 PM
Author: irradiated organic girlfriend

it's like dialing dominoes and pizza hut on a three-way call and watching the two pizza places be like "no man you're calling me"

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515145)



Reply Favorite

Date: December 16th, 2025 7:08 PM
Author: appetizing stain theatre

i'd try gemini, i thought chatgpt was giving the softest, most PC answer to everything. chatgpt used to be ahead but when they had everyone making friends with the AI it seems like they moved too far the other direction and totally kneecapped it

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515142)



Reply Favorite

Date: December 16th, 2025 7:13 PM
Author: pea-brained cerebral locale newt

its too safety coded right now for sure. but I'm mostly using it for machine learning and math, and 5.2 gives me the best answers for that. gemini has its strengths, but haven't been using it, because the api isn't on azure. I might make a new google account actually so I can get free console credits again and then I'll start using it.

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515158)



Reply Favorite

Date: December 16th, 2025 7:20 PM
Author: appetizing stain theatre

totally fair. i'm not using it for machine learning and math at all. my use cases are brainstorming for restaurants and stores around me, gym routine/personal trainer stuff, trip planning for sights, messages on dating apps,and interpersonal advice at work

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515194)



Reply Favorite

Date: December 16th, 2025 7:50 PM
Author: irradiated organic girlfriend

The only reason I use 5.2 is for validating answers from my own LLMs.

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515288)



Reply Favorite

Date: December 16th, 2025 7:50 PM
Author: Orange doobsian antidepressant drug jap

It's so ass

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515289)



Reply Favorite

Date: December 16th, 2025 8:11 PM
Author: pea-brained cerebral locale newt

I suppose it depends on what you are using it for and also the intelligence of the user. It is super human on most narrow to semi-general tasks. For me that is good enough to make my abilities post-human.

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515339)



Reply Favorite

Date: December 16th, 2025 8:18 PM
Author: Orange doobsian antidepressant drug jap

Coding a discord replacement in Assembly language

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515372)



Reply Favorite

Date: December 16th, 2025 8:31 PM
Author: pea-brained cerebral locale newt

lol jesus fuck that

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515418)



Reply Favorite

Date: December 16th, 2025 8:32 PM
Author: soul-stirring harsh toilet seat

Why?

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515421)



Reply Favorite

Date: December 16th, 2025 8:53 PM
Author: Orange doobsian antidepressant drug jap

Performant

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515470)



Reply Favorite

Date: December 16th, 2025 8:25 PM
Author: soul-stirring harsh toilet seat

What are you nerds even using this Chinese scam garbage for?

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515403)



Reply Favorite

Date: December 16th, 2025 8:32 PM
Author: pea-brained cerebral locale newt

I'm jewish

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515420)



Reply Favorite

Date: December 16th, 2025 11:34 PM
Author: puce theater

Python can do a lot and with AI you basically have an on demand python writing gadget (in addition to everything else AI can do). I'd reverse the question and ask what your life is like if you DON'T have a use for this resource.

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515726)



Reply Favorite

Date: December 16th, 2025 11:41 PM
Author: low-t mewling stage faggotry

It gave me an actually good legal memo yesterday (I.e., it largely held up after reviewing the sources—which actually existed) after a real rough patch of consistent SPS.

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49515733)



Reply Favorite

Date: December 18th, 2025 9:37 AM
Author: pea-brained cerebral locale newt

One of the things that I like about it is it is much more epistemically conservative and skeptical, which seems like it would be useful in correcting the failure modes it has in law (citing fake stuff confidently).

(http://www.autoadmit.com/thread.php?thread_id=5811106&forum_id=2)#49518890)