Mode, depth, and parallelization now asked in one call (3 questions) instead of 3 separate interactions.