> Me too. Claude Opus 5's English is insufferable. Opus 4.6-4.8 was more reasonable.

I suspect this degradation is happening because the AI labs are using the LLM's output to feedback into the input, to create a thinking loop, and they're optimizing that.