GSoC 2026 AI Mentor for PocketCode Students
44 views
Skip to first unread message
Zayden Zhu
Mar 4, 2026, 3:56:51 AMMar 4
to catrobat
Hi everyone,
My name is Zikun Zhu and I also go by Zayden. I am a graduate student at Stanford Graduate School of Education studying Learning Design and Technology, and a software engineer on Code in Place, Stanford's free CS education initiative. Before that I was full-time teaching staff at the National University of Singapore, teaching data structures and algorithms, software engineering, and machine learning courses to undergraduates.
The AI Mentor project stood out to me right away. The pattern I saw over and over in my courses was students struggling with debugging not because they lacked ability but because they lacked a patient guide at the exact moment of confusion. That is exactly what this project builds. The AI Tutor SDK has a clean foundation. The PromptStrategy interface shows thoughtful iteration on how to frame the conversation between student and AI, and the Intent-based approach through AiAppLauncher lets users leverage AI apps they already trust. I have been reading through CatroidApplication.java and the Koin setup in CatroidKoinHelper.kt to understand the integration surface. There are dependency alignment questions to work through, but the path forward looks tractable.
What draws me most is the chance to think about prompt design not just as engineering but as pedagogy. The difference between a prompt that gives a student the answer and one that nudges them toward discovering it themselves is exactly the kind of problem my LDT training prepares me for.
I am setting up the development environment and exploring the open issues. My GitHub is https://github.com/zikunz.
Zayden
My name is Zikun Zhu and I also go by Zayden. I am a graduate student at Stanford Graduate School of Education studying Learning Design and Technology, and a software engineer on Code in Place, Stanford's free CS education initiative. Before that I was full-time teaching staff at the National University of Singapore, teaching data structures and algorithms, software engineering, and machine learning courses to undergraduates.
The AI Mentor project stood out to me right away. The pattern I saw over and over in my courses was students struggling with debugging not because they lacked ability but because they lacked a patient guide at the exact moment of confusion. That is exactly what this project builds. The AI Tutor SDK has a clean foundation. The PromptStrategy interface shows thoughtful iteration on how to frame the conversation between student and AI, and the Intent-based approach through AiAppLauncher lets users leverage AI apps they already trust. I have been reading through CatroidApplication.java and the Koin setup in CatroidKoinHelper.kt to understand the integration surface. There are dependency alignment questions to work through, but the path forward looks tractable.
What draws me most is the chance to think about prompt design not just as engineering but as pedagogy. The difference between a prompt that gives a student the answer and one that nudges them toward discovering it themselves is exactly the kind of problem my LDT training prepares me for.
I am setting up the development environment and exploring the open issues. My GitHub is https://github.com/zikunz.
Zayden
Reply all
Reply to author
Forward
0 new messages