At least when I saw this thread I ran subset of my evals and therewasn’t any degradation. But I am using opencode with business copilot so it might be completely independent from average code + Claude subscription experience.