LEGO Mindstorms robots are a popular educational tool for teaching programming concepts to young learners. However, learners working with these robots often lack sufficient feedback on their programs, which makes it difficult for them to reflect on domain concepts and may decrease their motivation. We see an opportunity to introduce feedback into LEGO Mindstorms programming environments by having the robot itself deliver feedback, leveraging research on learning companions to transform the programmable robot into a social actor. Our robot, ROBIN, provides learners with automated reflection prompts based on a domain model and the student’s current program, along with social encouragement based on a theory of instructional immediacy. We hypothesize that by having the robot itself provide cognitive and social feedback, students will both reflect more on their misconceptions and persist more with the activity. This paper describes the design and implementation of ROBIN and discusses how this approach can benefit students.