To date, much of the related scholarly activity related to ChatGPT in educational contexts has been focused on the broad promises and perils of the tool. While some studies have explored students’ self-reported use of ChatGPT for programming assistance, little empirical research exists regarding how students use ChatGPT to learn to program. As an experienced educator but novice programmer, I address this research gap by conducting an autoethnographic study using ChatGPT to learn to program in Python. Over five learning sessions, I engaged with ChatGPT-4o to work through course materials from a first-year undergraduate programming class for non-computer science students. During those sessions, I fluctuated between my roles as experienced educator and novice programmer (student), applying my prior teaching experience and understanding of programming education to evaluate and adapt my interactions with ChatGPT. Overall, ChatGPT was able to generate pedagogically sound tutoring sessions that included experiential learning activities. These were a good starting point, but ChatGPT lacked the ability to autonomously assess my understanding and dynamically adapt the tutoring sessions or to my understanding and skills. Thus, I had to rely on self-monitoring to decide when to progress or ask for remediation. Additionally, ChatGPT’s explanations were often too long or inappropriate for my level which hindered my learning process. I end by evaluating my experiences with ChatGPT through a sociocultural lens and suggest that ChatGPT is best conceptualized as a (limited) learning-mediating tool rather than a more knowledgeable other.