The extension provides confidence scores for reasoning which are simple to understand and provides a timeline view of the scores. The extension also warns when the model responses degrade in reasoning.