Hello, I work on many projects, and lately I’ve found myself unable to justify using Claude 4 Sonnet—both because of its cost and its declining performance. Honestly, it feels like it’s gotten dumber over time. Even with carefully crafted prompts and clear context, I often have to repeat the same instruction six times before it does what I want.
In contrast, Claude 3.7 handles the same task correctly on the first try.
Isn’t that a bit absurd?