Date: April 4th, 2026 10:38 AM
Author: chilmata
My personal $200/month Max Claude was dumber than my $200/month business plan Claude and that was immediately noticeable with my first few prompts.
But I just saw some guy claim he saw in the leaked Claude code that they have purposely dumbed down Claude.
Another guy found a cache bug. People are MAF raging about how Claude has nerfed usage limits and they are running out of usage much faster than normal. I'm hitting usage limits on Claude much faster as well and Anthropic has not been transparent about the problem.
The cache bug basically overcharges you tokens to review every single line of context in your chat or project if Claude has been idle for over an hour.
So if you constantly use Claude you will run out of usage and if you don't use Claude constantly you will run out of usage just as fast due to the cache bug.
Dumbing down the model means Claude makes more mistakes keeping your usage up.
(http://www.autoadmit.com/thread.php?thread_id=5853409&forum_id=2)#49793117)