AI

AI Class Notes #1: When to Struggle

Note: This is the start of a new series that will appear on a regular basis. I'd love to hear from you and find out what you think about the Gen AI school push. I promise only one per day. - Miguel Guhlin

You've probably noticed the conversation around AI in schools is shifting. It's no longer just about whether students are using it—it's about whether they're learning to think alongside it. A former OpenAI researcher says detection is a lost cause. A writing professor argues the real skill is knowing when not to use AI. And new research shows AI-generated summaries still can't be trusted for academic work. Whether you're rethinking assignments or exploring new tools, these updates will help you make more informed choices. The 3-2-1 at the end wraps things up for you. Also, be sure to check the TCEA TechNotes blog for upcoming entries on this subject.

Headlines

The debate over AI in schools is shifting from "how do we catch cheaters?" to "how do we teach judgment?"

• Karpathy on Detection. Former OpenAI researcher says AI detectors are "doomed to fail." Discuss: Should schools stop trying to detect AI use and focus on assessment redesign?
• China's AI Education Plan. Ministry aims to bring AI to all K-12 schools by 2030, including facial monitoring. Discuss: What are the tradeoffs between personalization and student privacy?
• AI Summaries Fall Short. Study finds even top tools achieve under 70% accuracy and often exaggerate findings. Discuss: When should students trust AI-generated summaries—and when shouldn't they?

Perspectives

Educators and researchers are converging on a theme: AI literacy isn't about the tool—it's about thinking.

• Teaching When to Struggle. Writing professor shifts from neutral observer to guide with a point of view. Try it: Ask students to draft with and without AI, then justify their choices aloud.
• AI Literacy as Critical Practice. Essay argues true literacy means knowing when not to use AI. Try it: Run an "unplugged" classification activity to teach how AI sorts information.
• How Educators Use Claude. Anthropic report finds curriculum development is the top use; grading remains contentious. Try it: Reflect on your own AI use—where do you collaborate vs. delegate?

Tools

These platforms are some that people have mentioned to me. They are designed with educators in mind. Each offers a different entry point depending on your needs. There are many others. What are some you would recommend? Reply to this post with your own use cases and I will include them in future issues.

• MagicSchool. 80+ tools for lesson plans, rubrics, and student AI literacy. Try it: Generate a quiz draft, then refine it with your own questions.
• Brisk Teaching. Chrome extension for differentiation and feedback inside Google Docs. Try it: Use it to adapt a reading passage for multiple reading levels.
• NotebookLM. Google's research assistant with source-grounded summaries and citations. Try it: Upload a unit's readings and generate a study guide for review.
• BoodleBox. Collaborative AI platform with multiple models and custom bot creation. Try it: Build a custom bot to help students analyze primary sources.
• Colossyan. AI video creator with avatars, quizzes, and branching scenarios. Try it: Create a short instructional video for a flipped classroom lesson.

3-2-1 Reflect

3 things to consider:

• AI detection may be a losing battle—what does that mean for how you design assessments?
• Knowing when not to use AI might be the most important skill you can teach.
• AI-generated summaries still miss the mark—verification remains essential.

2 questions to ask yourself:

• Where in your practice do you collaborate with AI vs. delegate to it?
• How are you helping students develop judgment about when to struggle on their own?

1 thing to try:

• Pick one tool from the list above and explore it for 15 minutes. What's one task it could take off your plate? ---# AI Class Notes #2: Policy First, Play Later

One of the big challenges for schools, nonprofits, and businesses? If you don't have a plan, policy, or set of guardrails governing how you use Gen AI, you may be sticking your head in the sand. Whether you support Gen AI use or not, you still need a plan for dealing with it. Gen AI tools are quickly becoming ubiquitous, and you need agreement on how to use them—explore one approach via the Gen AI Adoption Checklist for K-16 Educational Institutions (Note: Password in comments on this post).

Headlines

Google is making moves, OpenAI is adjusting pricing, and the debate over AI in classrooms continues to heat up.

• Gemini Gets Notebooks. Organize chats and files with NotebookLM sync. Discuss: How might organized AI workspaces change student research habits?

• ChatGPT Pro at $100. New mid-tier plan bridges $20 and $200 options. Discuss: Should districts budget for premium AI access, or stick with free tiers?

• Gemma 4 Released. Google's most capable open models for advanced tasks. Discuss: What role should open-source AI play in school technology decisions?

• Pause on Classroom AI?. Op-ed calls for moratorium until independent research catches up. Discuss: What evidence would you need before expanding or pausing AI use in your school?

Perspectives

Voices from higher ed and industry are weighing in on how to use AI responsibly—and what happens when you don't.

• AI & Academic Integrity. Cornell's guide to writing with AI authentically. Try it: Have students compare their voice in AI-assisted vs. unassisted drafts.

• Enterprise AI ROI Benchmarks. 2026 data on productivity gains and adoption trends. Try it: Use the benchmarks to frame a faculty discussion on measuring AI impact.

Tools

From dashboards to flyers to 3D visualizations, these tools put Gen AI to practical use.

• KPI Dashboards with Gen AI. Build interactive school dashboards without coding. Try it: Create a dashboard tracking attendance or assignment completion for your class.

• News-Style Pet Flyers. Teach real-world writing through missing pet scenarios. Try it: Have students write a news-style flyer for a classroom mascot or lost item.

• Vibe Code for Google Sheets. Connect custom GPTs to spreadsheets for quiz data. Try it: Build a simple quiz bot that logs student responses to a shared sheet.

• Gemini 3D Visualizations. Ask Gemini to visualize concepts with interactive models. Try it: Prompt Gemini to "help me visualize how photosynthesis works" for a science lesson.

• Google AI Edge Eloquent. On-device AI assistant app from Google. Try it: Test offline AI capabilities for classrooms with limited connectivity.

• Pika AI Self. Create a personalized AI agent as a digital extension. Try it: Explore with students how identity and voice translate into AI personas.

3-2-1 Reflect

3 things to consider:

• Having no AI policy is itself a policy—one that leaves decisions to chance.
• Google's notebook and visualization features are blurring the line between chatbot and research tool.
• The call for an AI pause reflects real concerns about moving faster than evidence allows.

2 questions to ask yourself:

• What would a responsible AI adoption checklist look like for your specific context?
• Are you using AI tools to replace thinking, or to extend it?

1 thing to try:

• Draf