The most prestigious law school admissions discussion board in the world.
|
 |
 
|
Anyone still in denial that GPT 5.2 is smarter than any of us?
Poast new message in this thread
Date: January 18th, 2026 8:38 AM Author: appetizing green home toaster Now that 5.2 pro solved multiple erdos problems with minimal prompting? (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598279) |
 |
Date: January 18th, 2026 8:43 AM Author: Scarlet national what is erdos problem
|
|
Date: January 18th, 2026 8:47 AM Author: Mischievous trailer park mad-dog skullcap Engineering (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598290) |
|
Date: January 18th, 2026 9:08 AM Author: appetizing green home toaster erdos was a Jewish math prodigy who defined hundreds of open problems in pure math (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598312) |
Date: January 18th, 2026 8:47 AM Author: Mischievous trailer park mad-dog skullcap It clearly is but it can’t do shit. Like ok start a brand new company with no white collar employees and make a better cheaper Car or Console. Or tell me exactly when to buy a stonk. Best thing it can do rn is allow High Losers to Scale. Claude CoWork, named after CoWgod, is a breakthrough imo. (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598288) |
|
Date: January 18th, 2026 9:06 AM Author: appetizing green home toaster allowing high losers and mid nerds to Scale is YUGE though. basically a paradigm shift (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598310) |
|
Date: January 18th, 2026 9:26 AM Author: Mischievous trailer park mad-dog skullcap Please help me scale. I make 0k (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598332) |
|
Date: January 18th, 2026 9:48 AM Author: appetizing green home toaster I don't actually know what Scale means (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598371) |
|
Date: January 18th, 2026 9:53 AM Author: Mischievous trailer park mad-dog skullcap It's like Leverage. just a word to use (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598384) |
|
Date: January 18th, 2026 2:57 PM Author: supple mental disorder (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598948) |
|
Date: January 18th, 2026 3:28 PM Author: Exciting sickened immigrant coffee pot I'm scaling, daddy! (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49599006) |
|
Date: January 19th, 2026 10:51 AM Author: Cordovan legend native (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49600792) |
|
Date: March 9th, 2026 10:32 PM Author: which is what makes time travel possible (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49730886) |
|
Date: January 18th, 2026 9:11 AM Author: Ruby big weed whacker You people are so fucking retarded. (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598314) |
|
Date: January 18th, 2026 9:12 AM Author: appetizing green home toaster I'm Jewish autistic and gay (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598315) |
|
Date: January 18th, 2026 9:48 AM Author: chest-beating locus (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598374) |
|
Date: January 18th, 2026 9:49 AM Author: appetizing green home toaster yeah they are done here (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598375) |
|
Date: January 18th, 2026 10:02 AM Author: Stubborn Degenerate I use it for codecucking, although I write much less code now as a manager. It saves me time but needs my oversight because its first few suggestions are sometimes wrong, hard to understand, not following specific conventions, or some other reason. Eventually it does get things right, but for the code I work on it needs human oversight by an experienced engineer; otherwise, it'll turn the codebase into a mess. (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598398) |
|
Date: January 18th, 2026 11:53 AM Author: appetizing green home toaster Needs oversight by a human -- yes. But an intelligent human with gpt 5.2 is immeasurably better than one without. With this latest version -- 5.2 -- I honestly don't think its "wrong" that often. If you feed 100K tokens in code it lazy reads like half of it so you just need to continually give it more and more context either from the codebase or outputs and it converges on the answer. Or if you already directionally know where the problem is coming from even easier. Whats powerful to me about it isn't getting a single correct answer right away, but the fact that you can skip huge amounts of detail and time. It becomes all about high level orchestration. (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598589) |
|
Date: January 18th, 2026 1:25 PM Author: Stubborn Degenerate Yes, I'm aware that this is possible. I just don't write much code these days and therefore haven't explored AI fully because I spend most of my time doing pointless manager work. (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598781) |
|
Date: January 18th, 2026 11:54 AM Author: Dun state turdskin Sup Sam Altman ? (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598591) |
|
Date: January 18th, 2026 12:03 PM Author: appetizing green home toaster To me smarter just means superhuman in intelligence. I don't think it needs to be replace full human jobs as autonomous intentional agents. The fact that it can easily max out the lsat as a warmup exercise and solve open problems in mathematics that stumped humanity for decades makes it smarter. It is functionally smarter at intelligence and reasoning based tasks. (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598611) |
|
Date: January 18th, 2026 1:28 PM Author: Swashbuckling jet-lagged internal respiration it gets accounting questions wrong all the time (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598790) |
|
Date: January 18th, 2026 4:04 PM Author: appetizing green home toaster yeah. accounting and law are human made culturally and historically and socially contingent systems with a huge space of alternate plausible possibilities of how it could have been done. AI is better at predicting underlying structure, not just regurgitating systems already in place. (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49599107) |
|
Date: January 19th, 2026 8:54 AM Author: Scarlet national isnt accoutning just in-out=accumulaton with T accounts where u memorize asset side is debit for inc and equity + liability is credit for inc (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49600570) |
|
Date: January 19th, 2026 9:30 AM Author: appetizing green home toaster Was assuming he was talking about the audit and regulatory side of it. The bookkeeping part has been automated since the invention of basic computation. You really think something that is solving Erdos problems cant do that? Reason it’s not good at big 4 accounting like audit and tax is it needs to reproduce bullshit arbitrary human regulations perfectly or it’s a liability |
|
Date: January 19th, 2026 10:14 AM Author: Scarlet national is this all because of Sarbane oxley? what does a auditor do on the regulations? (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49600705) |
|
Date: January 19th, 2026 10:28 AM Author: Swashbuckling jet-lagged internal respiration Sox did not. The regulations are principals based, not rules based. (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49600735) |
|
Date: January 19th, 2026 11:37 AM Author: appetizing green home toaster I’d say it’s a mix between principle and rules based but principle based can still be arbitrary and contingent because the space of possible “underlying principles”you can use is enormous and even abstract principles can never be truly universally applied. Humans learn it based on social and professional conventions. (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49600927) |
|
Date: January 19th, 2026 10:28 AM Author: Swashbuckling jet-lagged internal respiration There is still some art to it as the rules in the US are principal based and not rules based like IRFS. Chatgpt is still weak at fuck at making judgements in accounting. (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49600733) |
|
Date: January 19th, 2026 11:40 AM Author: appetizing green home toaster Yeah it will get there. Chat GPT is pretty weak in general when it comes to making strong judgements about things that have real world consequences and most of that is because they sanded down that stuff due to liability. They don’t want it doing things that look like proto intentionality or people making decisions that have real world consequences. For fields like law and accounting they might end up needing a different kind of product than whats given to the general public (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49600945) |
|
Date: January 18th, 2026 1:36 PM Author: pale dopamine piazza It’s still shit at chess, doesn’t seem capable of remembering the board (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49598809) |
|
Date: January 19th, 2026 9:41 AM Author: Floppy new version My job is solving erdos problems so I'm starting to get a lil worried (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49600634) |
|
Date: January 19th, 2026 11:41 AM Author: appetizing green home toaster 180 (http://www.autoadmit.com/thread.php?thread_id=5823517&forum_id=2Vannesa#49600948) |