I co-teach AP Computer Science A through Microsoft’s TEALS program. The classroom runs on Chromebooks, Google Classroom, and code.org (AWS). Corporate infrastructure top to bottom. This year I added an AI tutor. That’s apparently the controversial part.
The research is interesting: a Wharton study found students using standard ChatGPT performed 17% worse on exams—the “crutch” effect. But students using AI with pedagogical guardrails showed no negative effect. The problem isn’t AI in education. It’s unguided AI. So I built a tutor that asks probing questions instead of giving answers. I’m sharing the prompt I use and how to set one up yourself.
While, China made AI education mandatory for six-year-olds this year. We’re still deciding whether to block ChatGPT.
Hard pass. There absolutely should be no AI in any classroom under any circumstances. The whole point of a classroom is to build a foundation on which to understand the fundamentals before they slap a set of training wheels on and vibe-code their way into disaster. Most of these LLMs ignore whatever guardrails you slap on them far too frequently.
The most important lesson these kids need to learn is if you can’t do it yourself, you shouldn’t be letting an LLM do it for you. If the best you can say about the effects is “This version doesn’t seem to be actively harming them” then the bar is in hell, and we shouldn’t be playing with these tools at all at this point.