yep, agree. One big part of the experiment was to see how well it does the reasoning by asking it to output the reasoning.