I recently switched over to anthropic models, and they seem to have a better “attitude” compared to openai.
I also like the (marginally) fewer hallucinations, but it means more refusals (when uncertain) in API calls. Looking forward to trying out this “step change” when it’s released.
I find it helpful to switch the default diff2 conflict style (which just shows the two chamges) to diff3 (which shows the unchanged lines/common ancestor too).