The most prestigious law school admissions discussion board in the world.
|
 |
 
|
GPT 5.2 is again so far ahead of everything that its a joke
Poast new message in this thread
Date: January 12th, 2026 2:22 PM Author: chopped unc I was honestly wondering if Claude was going to jump ahead with 5.1 being kind of lackluster. but then 5.2 made a huge leap ahead of opus. and obviously gemini is still way behind both of those (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49583695) |
 |
Date: January 12th, 2026 2:23 PM Author: #notflame how can you tell? (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49583697) |
|
Date: January 12th, 2026 2:24 PM Author: zesty I'm just guessing (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49583701) |
|
Date: January 12th, 2026 3:10 PM Author: Nazca Redlines (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49583849) |
|
Date: January 12th, 2026 4:32 PM Author: chopped unc (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584035) |
Date: January 12th, 2026 2:27 PM It's a hilarious/brutal market. There's effectively no lock-in, so I just switch between whoever is good right now. Think they're the current winner, I never need to switch bots because it can't do something. In 4 the python became passable, now in 5.2 it's really good. I never bother coding "by hand" any more. (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49583708) |
|
Date: January 12th, 2026 2:31 PM Author: ~~(> ' ' )> I've used ChatGPT a few times to make scripts for Windows, and it always fucks up. Is it better at real coding? (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49583728) |
|
Date: January 12th, 2026 2:32 PM Windows Scripts? You mean PowerShell? I haven't tried that. It does just fine with python, which isn't surprising. Python is one of the most documented languages on the planet. (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49583733) |
|
Date: January 12th, 2026 2:37 PM Author: ~~(> ' ' )> Powershell and just good old batch files for cmd. (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49583753) |
|
Date: January 12th, 2026 4:42 PM Author: chopped unc i dont believe for a second that it is messing up powershell scripts. maybe a script 3K lines long you glued together to train a model might not work the first go around. (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584053) |
|
Date: January 12th, 2026 4:58 PM Author: zesty (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584090) |
|
Date: January 12th, 2026 6:07 PM Author: ~~(> ' ' )> Nothing nearly that long. And it has been things like CLI syntax errors and mistakes in handling errors. (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584273) |
|
Date: January 12th, 2026 6:15 PM Author: chopped unc well idk. i dont use powershell I use ubuntu and mac os only. it never fucks up bash and python scripts for me. the older versions used to fuck them up a lot. I haven't seen gpt make a syntax error since like gpt 4.1. maybe once in a while on o3 and o4. but not since it switched to 5. (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584292) |
|
Date: January 12th, 2026 2:34 PM Author: fully online and responsive No it's horrible (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49583738) |
|
Date: January 12th, 2026 2:41 PM Author: ~~(> ' ' )> One of the things I've found most frustrating is that when you're telling it to fix an error (which is annoying enough), it will often change (and break) other things that aren't at issue and were working just fine. It will even do that sometimes when you specifically tell it not to touch anything else. (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49583773) |
|
Date: January 12th, 2026 2:45 PM What are you trying to do and what prompt did you give the bot (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49583786) |
|
Date: January 12th, 2026 2:55 PM Author: fully online and responsive I am coding a discord replacement in assembly language
But it can't even give the right proxmox commands. It always says pct restart instead of pct reboot for instance, It's bad |
|
Date: January 12th, 2026 4:34 PM Author: chopped unc lol yeah no wonder. you are using it to code assembly language. its really good at the popular languages. python is probably its best. also you can't always just give it some abstract prompt and expect it to guess the "right" implementation you have in mind. you need to be feeding it real concrete evidence from outputs etc. (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584036) |
|
Date: January 12th, 2026 5:19 PM (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584139) |
|
Date: January 12th, 2026 5:20 PM Author: As far as they will go but even farther (π§) were u possessed by NSAM (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584143) |
|
Date: January 12th, 2026 5:50 PM Author: fully online and responsive I'm flaming online (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584236) |
|
Date: January 12th, 2026 6:09 PM Author: https://i.imgur.com/ovcBe0z.png (mail@baumeisterlaw.com) (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584278) |
|
Date: January 12th, 2026 3:01 PM Author: zesty Scripts are coding and it's good also try codex (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49583832) |
|
Date: January 12th, 2026 3:10 PM Author: theranchman (bizarro world where opus 4.5 doesn't exist) (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49583847) |
|
Date: January 12th, 2026 4:34 PM Author: Mainlining the $ecret Truth of the Univer$e (One Year Performance 1978-1979 (Cage Piece) (Awfully coy u are)) (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584037) |
|
Date: January 12th, 2026 4:35 PM Author: chopped unc its much better than opus 4.5. opus 4.5 just comes up with all of this stuff that literally doesn't matter. like give it something thats broken and it will give you 100 edge cases that matter in 1 in a billion situations and tell you to fix them. gpt 5.2 gives you useful stuff every time. opus 4.5 can be ok though if you have gpt 5.2 to keep it in check but still nowhere even close to as good at coding and math as gpt 5.2. also gpt 5.2 is much more conservative about conclusions. opus 4.5 will be super super certain about an answer when its just guessing. the guess might be correct like 1 in 5 times but the times its hallucinating add up. there are probably things opus is better at like writing and business intelligence. but im 100% sure gpt 5.2 is substantially better at coding and math because the gap is big. (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584039) |
|
Date: January 12th, 2026 4:54 PM Author: Non sequitur ChatGPT is smarter likexo autist
Claude is more polished like legally blonde or whatever |
|
Date: January 12th, 2026 5:46 PM Author: chopped unc yeah i agree with that (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584221) |
|
Date: January 12th, 2026 8:03 PM Author: Mainlining the $ecret Truth of the Univer$e (One Year Performance 1978-1979 (Cage Piece) (Awfully coy u are)) (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584733) |
|
Date: January 12th, 2026 4:43 PM Author: Taylor Swift is not a hobby she is a lifestyle (πΊπΈ π΅π±) (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584058) |
|
Date: January 12th, 2026 5:04 PM Author: cowgod What a Joke (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584107) |
|
Date: January 12th, 2026 5:46 PM Author: chopped unc (http://www.autoadmit.com/thread.php?thread_id=5821130&forum_id=2Reputation#49584223) |