Bottom line: More and more AI companies say their models can reason. Two recent studies say otherwise. When asked to show their logic, most models flub the task – proving they're not reasoning so much ...