Opus 5.5 still makes mistakes all the time even if the prompt isn't super complex. A stronger model make mistakes less often, thus wasting less of my time.