The use of robots to teach students with autism spectrum disorder communication skills has basis in the literature; however, research investigating the effects of teaching coding or programming of robotics to promote learning in STEM to this population has not yet been conducted. The purpose of the current study was to evaluate the effects of teaching one code explicitly, using model-lead-test on the following dependent variables: (a) acquisition of the explicitly-taught code (i.e., robotic movement); (b) generalization of the explicitly-taught code to other novel codes (i.e., robotic sounds, light effects, complex movements), and (c) self-generated novel sets of codes. Results of the multiple probe across participants design demonstrate that all three students with ASD and challenging behaviors were able to acquire the initial code, generalize the initial code to novel codes, and self-generate (i.e., create, test, and evaluate) their own coding. Implications for practitioners, study limitations, and recommendations for future research are discussed.