I’ve started writing this bit of the newsletter three times now, and three times I’ve had to start over. This being the fourth incarnation of a brief introduction to this week’s things I thought you might like, I’m sitting at my desk in this comfy big-butt chair, mulling over the already inordinate amount of AI tool usage to make the stupid, mundane, completely unnecessary tasks we are obliged to do every day in whatever industry we work in easier to accomplish without ever asking the question, “why are we doing this anyway?”
I have no qualms about using AI tools. If they’re the right tool for the job, use them. But always question what the job is, why you’re doing it, and whether you should keep doing it.
Just because you can do something because it’s easier to do with AI help doesn’t mean it was ever worth doing in the first place.
Anyways, on with this week’s 10 things…
10 Things - 11 September 2026
We Can Limit AI Without Leaving Kids Unprepared
New York just drew a hard line on student-facing AI through eighth grade, and Andrew Marcinek’s follow-up asks the question the moratorium leaves open: what should kids learn about AI while it’s off limits, and how should that change once it’s allowed? He pulls apart a PNAS math study where unguided chatbot use left students worse off later, against a Harvard physics trial where a carefully designed AI tutor outperformed in-person teaching — and lands on a line worth sitting with: when the assignment’s done and the tool’s closed, the student still has to be able to think.
No More Muddling Through
Mark Humphries thinks the quiet truce between students and professors over AI is about to break. This fall’s incoming class used AI throughout all four years of high school, and the models themselves can now replicate a full research process end to end — which means the avoidance strategy so many instructors have relied on stops working right about now.
Whose Literacies Are Shaping Your Portrait of a Graduate?
Angela Stockman’s working with a New York district on Portrait of a Graduate design, and she’s noticed something most of these processes skip: nobody stops to ask what “literacy” even means before building the whole plan around it. Her protocol for surfacing what a community doesn’t yet know it values is worth stealing for any vision-setting work.
Five Great Questions For “Evidence-Based” Programs
Buried at the bottom of an EdWeek guest post on the “science of math,” Rene Grimes drops five questions that Peter Greene thinks should be taped above every curriculum adoption meeting: what did the original research actually investigate, which students and settings were studied, does the evidence support the recommendation or just an interpretation of it, what limitations should shape its use, and what would tell us it’s actually working here. Greene’s breakdown is a good gut check for the next AI tool pitch, too.
Get Curious Before You Get Busy
A semester coaching first-gen business students taught Brian LeDuc something worth remembering at the start of any school year: the sharpest insight into what people actually need rarely comes from the questions leadership already has. His reflection is a short case for listening before you plan.
They Invited Educators to See the Future of Work. It Was a Surveillance State.
A construction company invited teachers to preview its latest tech — AR headsets, exoskeletons, wearables that flag “extended periods of inactivity” — meant to excite them about trades careers. Instead, the piece argues the demo revealed a path toward deskilling the very workers it claims to train, concentrating judgment at the top and leaving less room for anyone to learn their way up. Worth reading before the next vendor demo of AI classroom tools.
Five Questions to Ask AI to Minimize Sycophancy
Jason Neiffer’s follow-up to his piece on AI’s flattery problem hands you five prompts built to push back against a model that just wants to agree with you — genuinely useful if you’re using AI for planning, feedback, or coaching prep and want something other than validation.
The Rise of AI Companions
A new Digital Future Center survey gets past the percentages and into what it actually feels like for people who’ve formed something like a relationship with an AI companion — trust, emotional connection, being understood, sometimes more than by the people around them. Some people are getting something real out of this, whatever we think about it personally.
The Power of Routine Creativity
John Spencer’s argument here is a useful correction for anyone chasing the big showcase project: the most powerful creative thinking in a classroom usually isn’t an event at all; it’s a habit built into ordinary, everyday work.
Economic Scenarios for Transformative AI
Tyler Cowen flags a new paper — with an Anthropic-heavy author list — that lays out sober, structured economic scenarios for AI’s impact between 2026 and 2030. If the daily AI-in-schools news cycle has you missing the forest for the trees, this is the forest.
That’s all, folks. Also, don’t forget what happened 25 years ago and how much worse the world has gotten since then.