I underscored using xhigh + fast mode when saying it’s so much better.

Now with Opus 4.7 of course the “burden” of adjusting reasoning effort has been taken away from you even at the API level.

In my experience people don’t change the thinking level at all.