Hey everyone,
Since the release of Sonnet 3.7 and 3.7 Thinking in Cursor, I’ve noticed a serious drop in output quality—not just from these models, but also from 3.5. The reasoning ability has deteriorated, and the model often fails to execute modifications correctly, sometimes getting stuck in a loop.
I used to rely on it for Flutter, Next.js, and React Native development, but now it struggles to function properly. It randomly generates errors, creates functions that make no sense, and constantly needs to reference the codebase, making it difficult to work efficiently.
Before the update, I could quickly develop well-structured, functional features within hours. Now, it’s practically unusable. For the time being, I’ll stick to coding manually and using Copilot instead.
Has anyone else experienced the same issues? Let me know your thoughts.