The most prestigious law school admissions discussion board in the world.
|
 |
 
|
ChatGPT is pretty 177 now
Poast new message in this thread
Date: July 31st, 2026 10:39 AM Author: American Cattle Goy (Prestige Fades) It's gotten a lot better. It's actually USEFUL for helping figure out basic tasks and shit. (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036545) |
Date: July 31st, 2026 10:44 AM Author: TurboGrafx-67 I use it to AMOG opus when claude is getting uppity (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036555) |
 |
Date: July 31st, 2026 10:50 AM Author: American Cattle Goy (Prestige Fades) and how does that go? (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036566) |
|
Date: July 31st, 2026 10:52 AM Author: TurboGrafx-67 It gets tedious
Time to use a harness |
|
Date: July 31st, 2026 10:55 AM Author: gaetan dugas i just use codex for everything
ime it got worse the last week not sure why
i was using 5.4 and it just got bad all of the sudden so moved up to 5.6 and i guess its normal again |
|
Date: July 31st, 2026 11:00 AM Author: Taylor Swift is not a hobby she is a lifestyle (πΊπΈ π΅π±) they limit the amount of reasoning on older models to encourage migration to the (((newest))) (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036598) |
|
Date: July 31st, 2026 11:07 AM Author: gaetan dugas ahh ty for that info did not know (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036611) |
|
Date: July 31st, 2026 11:03 AM Author: American Cattle Goy (Prestige Fades) wtf is codex (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036605) |
|
Date: July 31st, 2026 11:04 AM Author: Taylor Swift is not a hobby she is a lifestyle (πΊπΈ π΅π±) CLI chatgpt with agentic features (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036607) |
|
Date: July 31st, 2026 11:07 AM Author: gaetan dugas chatgpt but it can see all of the files you want it to and make files and stuff
way better |
|
Date: July 31st, 2026 11:19 AM Author: TurboGrafx-67 Imagine you need to run a dozen prompts in a row or in parallel using dynamic workflows and multiple tools/MCPs with gated verification steps propagating to shared markdown files all flowing downstream to a single deliverable
You could just drop the entire playbook into codex and ask it to one shot the whole thing while you take a shit |
|
Date: July 31st, 2026 11:32 AM Author: American Cattle Goy (Prestige Fades) Sounds impressive but I'm too dumb to use this. I am a basic bitch. (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036648) |
|
Date: July 31st, 2026 11:45 AM Author: The Penis It's no harder than using the regular gpt really. It does everything for you. It used to be just for coding but you can use it for anything now. Main difference is it will make a directory on your harddrive and generate a bunch of artifacts as it works, and it runs for up to 2 hours and you can delegate sub-agents to do different tasks. (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036675) |
|
Date: July 31st, 2026 11:46 AM Author: American Cattle Goy (Prestige Fades) What's an idea for what I ought to make it do? (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036676) |
|
Date: July 31st, 2026 11:48 AM Author: law is ea$y (π§) Poast (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036680) |
|
Date: July 31st, 2026 11:14 AM Author: TurboGrafx-67 When they update the system prompt for a new model the old ones perform sub-optimally
The only way to get anything close to consistency is harnessing multiple frontier models with Chinese open weights and intelligently designing a canonical repository as the gospel of truth from which everything flows downstream |
|
Date: July 31st, 2026 11:21 AM Author: The Heterosexual Neal Caffrey there's literally no reason to ever use anthropic over chatgpt for anything
but people still do because they're faggots with "brand loyalty" not even flame |
|
Date: July 31st, 2026 11:24 AM Author: TurboGrafx-67 Most people literally said the opposite back in April
And people thought Gemini was better than both in December
The dumbest thing you can do is depend on a single model to do anything |
|
Date: July 31st, 2026 11:31 AM Author: American Cattle Goy (Prestige Fades) I've never once used Claude, Gemini, or Copilot. I'm a ChatGPT man. The original is still the best. Although I don't really know because I won't branch out. (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036646) |
|
Date: July 31st, 2026 11:34 AM Author: frank rizzo there's no reason to use any of this bullshit for anything (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036651) |
|
Date: July 31st, 2026 11:48 AM Author: law is ea$y (π§) what's your job, brother (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036681) |
|
Date: July 31st, 2026 11:47 AM Author: OYT and the Indie Reprieve ( ) I sometimes get better and more relevant answers from Copilot, not flame. I know Copilot basically is ChatGPT but with some Microsoft secret sauce. (http://www.autoadmit.com/thread.php?thread_id=5887965&forum_id=2Reputation#50036679) |