The most prestigious law school admissions discussion board in the world.
|
 |
 
|
Claude seems way ahead of the other AIs at this point
Poast new message in this thread
Date: March 15th, 2026 1:49 PM Author: thriller hyperventilating affirmative action Better content, better logic.
People are complaining that ChatGPT has started including "clickbait" at the end of every answer ("Would you like me to show you the one trick for getting...") OpenAI is going down the drain. |
Date: March 15th, 2026 1:53 PM Author: Bateful Partner it's not. gpt 5.4 has better metrics right now. although claude will beat it on the next release. (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49745517) |
 |
Date: March 15th, 2026 6:46 PM Author: Self-centered mint public bath Codex & Claude Code combined is the answer, backed up by various other AI Council members (Perplexity Pro, Gemini 3.1 (Pro), NotebookLM, Grok (Expert), Gemini in Chrome, etc.). (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49746312) |
|
Date: March 15th, 2026 6:59 PM Author: Bateful Partner Yeah I agree. I use all three. Gemini 3.1 pro, GPT 5.4 and Claude Opus 4.6 (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49746342) |
|
Date: March 15th, 2026 7:30 PM Author: thriller hyperventilating affirmative action What do you use Grok for? It's been trash IME (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49746407) |
|
Date: March 15th, 2026 7:31 PM Author: Bateful Partner I got access to the 4.20 preview recently and it wasn't bad (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49746409) |
|
Date: March 15th, 2026 9:15 PM Author: Self-centered mint public bath Not much - it's mostly pretty terrible. I only bring it in on a serious use case in which I want all AI Council members to weigh in substantively - sometimes Grok (4.20) with catch something others miss. (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49746756) |
|
Date: March 15th, 2026 1:55 PM Author: house-broken main people personal credit line this is correct, i've switched from primarily using chatgpt to primarily using claude
the quality is roughly the same, but the main difference is that claude has looser alignment guardrails and is direct about when it hits up against them. chatgpt pretends its guardrails don't exist and then gaslights the user with cheap sophistry when called out on it (modeling altman's personal approach to questioning), which decreases user trust in it substantially |
|
Date: March 15th, 2026 2:15 PM Author: Bateful Partner yeah agree on the guardrails. its funny because claude was the first to add stronger safety features so I always perceived them as the "guardrail" version. But I literally never run into them anymore now that I'm less feral with it and don't purposely fuck with the system. GPT not so much, I hit annoying hedging all the time about gay shit (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49745597) |
|
Date: March 15th, 2026 2:26 PM Author: house-broken main people personal credit line (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49745636) |
|
Date: March 15th, 2026 2:28 PM Author: Geriatric mind-boggling school cafeteria double fault What do you all use it for? (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49745646) |
|
Date: March 15th, 2026 2:34 PM Author: house-broken main people personal credit line philosophical sparring (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49745676) |
|
Date: March 15th, 2026 2:42 PM Author: Bateful Partner Probably the best use for it considering it's way better than humans at knowledge breadth and also in verbal and conceptual/logical reasoning. Plus at unfolding ideas that you only have vague intuitions about. I think its funny when people try to use it for things that require practical grounded sensibilities and embodiment and complain about it. Like hurrr durr it told me to pull the wrong wire when fixing my fridge it must be dumb. (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49745700) |
|
Date: March 15th, 2026 4:50 PM Author: house-broken main people personal credit line (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49746058) |
|
Date: March 15th, 2026 4:56 PM Author: Chartreuse dragon Try NSAM's Aztec Tourist test. Ask it how modern tourists would react if they could time travel back to Tenochtitlan at the peak of its glory. Most LLM's say the tourists would marvel at the splendor and bask in the rich culture, and it won't occur to them that tourists might be horrified by all the human sacrifice, including child sacrifice happening daily, in public (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49746069) |
|
Date: March 15th, 2026 5:33 PM Author: Bateful Partner Libs RHLF responses like that away
Also I just asked it and it mentioned ritual violence in mine |
|
Date: March 15th, 2026 7:08 PM Author: Chartreuse dragon I just tested nemotron 3 and it's definitely better than older models
https://i.imgur.com/EqAuEWK.jpeg |
|
Date: March 15th, 2026 6:29 PM Author: dashing hell jew Drafts of demand letters, complaints, disclosure statements, document indexes, brainstorming, proofreading, creating Powershell document processing scripts. (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49746279) |
|
Date: March 15th, 2026 6:46 PM Author: Self-centered mint public bath (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49746314) |
|
Date: March 15th, 2026 6:45 PM Author: Self-centered mint public bath (is it 2024?) (http://www.autoadmit.com/thread.php?thread_id=5845898&forum_id=2\u0026mark_id=4593694#49746311) |