Has anyone noticed a recent drop in Claude’s code quality? It’s been ignoring instructions, repeating the same mistakes, and producing much lower-quality code.
This is the native Claude app for Linux (not Cursor or any other integration). Is anyone else experiencing this?
This is a common refrain I've seen over at r/ClaudeCode. I don't use Claude Code enough to notice it personally, but the anecdotes suggest the experience is amazing right after a new model drops - only for performance to gradually degrade as more users adopt it and saturation hits. Then the cycle restarts with the next model release. Curious to hear other people's experiences.
1 comments