[Bug] Claude models exhibit increased hallucination and context loss after Sonnet 4.6 release

Bug Description
What the hell happened since the claude sonnet 4.6 release? it feels like both models forget things after a few minutes. I told it to read a script and do something similar. It overengineered the wrong parts, seems to not understand what a prompt limit is so it tried some odd chunking and all sorts of other issues. I use claude because it used to understand what I wrote better than gpt. Now it feels more like gemini and hallucinating or overanalyzing/overengineering (doing way more than I ask for causing even more issues) way too much. This has been somewhat of a problem before but it seems to have gone up to an eleven with the recent release.

Environment Info

• Platform: win32
• Terminal: windows-terminal
• Version: 2.1.39
• Feedback ID: 9d96ccaf-5984-48d7-85c7-ed932159f81f

Note: Content was truncated.