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

Had to rewrite my entire Claude backend because Opus 5 is such shit

Like 90% of its output was literal gibberish and anytime you...
TurboGrafx-67
  08/05/26
Why don't you just use 4.5 then
The Dick
  08/05/26
4.6 is better; just saying 4.5 was good enough and the glory...
TurboGrafx-67
  08/05/26


Poast new message in this thread



Reply Favorite

Date: August 5th, 2026 1:38 AM
Author: TurboGrafx-67

Like 90% of its output was literal gibberish and anytime you asked it to review anything it invented more problems than it fixed

After spending days updating all my system settings and md files for the new system prompt and published anthro guidance, it finally doesn't suck

Annoying because I'm sure it will break again soon and never be as good as the 4.5 days when we were spoiled with virtually unlimited compute. The AI was already smart enough back then and everything worked fine

Now they're going to make it worse over time forever

(http://www.autoadmit.com/thread.php?thread_id=5889690&forum_id=2:#50046347)



Reply Favorite

Date: August 5th, 2026 1:48 AM
Author: The Dick

Why don't you just use 4.5 then

(http://www.autoadmit.com/thread.php?thread_id=5889690&forum_id=2:#50046350)



Reply Favorite

Date: August 5th, 2026 1:55 AM
Author: TurboGrafx-67

4.6 is better; just saying 4.5 was good enough and the glory days. And I had to rewrite shit because 4.6 was impacted too

My current workflow is still 4.6 max effort for all intensive tasks and 5 high / extra high to audit. Fable is reserved for high level architecture or areas that exceed my knowledge or cognition

But honestly Fable seems like a dumbass most of the time



(http://www.autoadmit.com/thread.php?thread_id=5889690&forum_id=2:#50046352)