News for hackers. Stuff that matters.
Re:Anthropic silently downgraded cache TTL from 1h → 5M on M... by sysid on Tuesday April 14, 2026 @04:48PM (#47771298)
Same experience, here. Very hard to base on facts, because every problem and prompt is an individual use-case and measuring agent reasoning quality is notoriously difficult anyway. But I spend a lot of time with Claude and my overall "feeling" fully matches your description. Quality has deteriorated, thinking takes longer and results become shallow. Something is off...

There are no comments on this comment yet.