Welcome new subscribers! If this is not your first issue of the newsletter, you may notice that it’s got a new name: Real-World AI for Teachers. The content and format haven’t changed, but the new name will help draw in more educators who may find this kind of information useful. As a reminder, this email is always free and will arrive in your inbox in the middle of every month. I’ll never sell you email address to anyone else, or send you any additional messages. But reach out to me if you have questions or want to dive deeper into any of the topics I share in these monthly newsletters.

NPR recently followed high school students from across the country as they created their own model AI policy for schools, tackling issues like bias, privacy, teacher autonomy, and responsible use.

What I like most is the shift in perspective: instead of only asking how to stop students from misusing AI, schools can also ask students what responsible use should look like. Giving them a voice in the conversation may build more ownership than another page of rules ever will.

Four San Antonio middle schools spent a year using an AI-powered reading and writing tool called Coursemojo. The results were encouraging in some classrooms and pretty flat in others.

That may be the most useful finding. AI tools don’t appear to be educational magic dust. Their value still depends heavily on how teachers weave them into good instruction. For schools rushing to buy AI tools, that feels like an important reminder: implementation matters at least as much as the tool itself.

A new look at Denver-area schools shows just how unsettled AI policy still is. Some districts are embracing classroom AI, while others are limiting access, especially for younger students.

The tension is revealing: many teachers worry AI could weaken critical thinking, yet most still want schools to teach students how to use it responsibly. That feels like the real challenge ahead—helping students gain AI fluency without giving away the thinking we want them to practice.

North Carolina has made AI professional development a statewide requirement. Under a new state law, the Friday Institute at NC State will create at least 10 hours of self-paced, tool-agnostic AI training, and teachers in traditional public schools, charter schools, and laboratory schools must complete it by June 30, 2028.

What caught my attention is what lawmakers actually want teachers to learn. The training will cover AI fundamentals and hallucinations, classroom uses like lesson planning and feedback, fact-checking AI output, privacy, bias, accessibility, and professional responsibility. It even specifically calls for training on “academic integrity and assessment redesign” as students gain greater access to AI.

That last part feels especially important. North Carolina is recognizing that preparing schools for AI requires more than learning which buttons to push. Teachers need time to think about how AI changes teaching, learning, and the way we ask students to demonstrate what they know.

Last month, I shared Defend the Delta, an idea from my upcoming book, The Learning Forge, designed to help students critically evaluate AI suggestions instead of simply accepting them.

This month, I want to move AI to a completely different part of the learning process: before students start the assignment.

The idea is pretty simple. Before students write the essay, build the model, create the presentation, design the experiment, or tackle whatever task you’ve given them, have them use AI to help answer one important question:

What would make this really good?

Then—and this is the important part—they have to decide whether the AI is right.

Teachers usually spend a lot of time deciding what quality looks like before an assignment ever reaches students. We build rubrics, write success criteria, create exemplars, and then hand all of that thinking over to them.

There’s an opportunity here to let students do some of that intellectual work themselves.

Give students the assignment before you give them the rubric. Have them ask an AI model to suggest the qualities of an excellent response. The AI might recommend things like strong evidence, clear explanations, accurate vocabulary, effective organization, creativity, or attention to the audience.

Students then have to evaluate that list.

Which suggestions actually matter for this assignment? Which are vague? What did the AI overlook? Is anything on the list there simply because it sounds like something that belongs on a rubric?

That last question can be surprisingly entertaining. AI loves a good generic rubric almost as much as teachers love a laminator.

I especially like doing this as a class conversation.

Students can generate possible criteria individually or in small groups and then bring their ideas together. As a class, we sort through the suggestions and decide what deserves to make the final list.

I might ask:

Eventually, we create our own class version of the success criteria or rubric.

The AI helped us brainstorm, but the students had to do the harder work of deciding what quality actually means.

You don’t need to redesign an assignment to try this.

Take something you already teach and simply hold back the rubric for a little while. Give students the task and ask them to use an AI model with a prompt like:

Then have students compare the AI’s suggestions with their own thinking.

You can even reveal your original rubric afterward and let students compare the two. Where did they agree? What did the teacher value that the AI missed? Did the students identify something worthwhile that wasn’t on the teacher’s rubric?

And if their ideas improve your rubric, change it.

I actually think that’s one of the best possible outcomes.

One of my big concerns about classroom AI use is that we tend to introduce it at exactly the point where students should be doing the most thinking: Now write the paragraph. Now solve the problem. Now generate the idea.

Moving AI earlier in the process opens up some much more interesting possibilities.

Instead of asking AI to create the work, students can use it as something to push against. They evaluate its suggestions, argue about quality, develop success criteria, and build a clearer picture of what they are trying to accomplish before they begin.

And once students have helped define what good work looks like, there’s one more useful step: give the criteria back to them at the end and ask them to evaluate their own work against the standards they helped create.

Now AI has supported the assignment twice, and neither time did it do the assignment for them.

That’s the kind of AI use I want more of in classrooms.

Want to see more? Check out the Learning Forge One-Pager for more ideas from the upcoming book.

The end of the summer is here! I’ve spent the past few months sharing the ideas in my upcoming book, “The Learning Forge” and getting TONS of useful feedback from fellow educators. The writing process has been exciting and I hope to be able to share more details in the coming months.

For the next month, though, it’s time to get nervous and excited for the start of another school year. I’m eager to put into practice the ideas that come out of the Learning Forge, and to share the results with you.

September’s newsletter will include some more lesson ideas that have been “field tested” in my very own 8th grade science classroom.

Paul (and the Codium Educational Consulting team)

P.S.

Don’t forget to email me with examples of how you’ve used the tools and strategies that I’ve shared.