The Law Is Already Moving
The question I open with this week is a ruminating concept for a presentation I have coming up soon.
“How many of you have a staff member who used AI to write a student document such as an IEP goal, a progress report, a recommendation letter in the last 30 days?”
A few hands. Maybe.
“How many of you know for certain they didn’t?”
Silence. The productive kind.
Here’s the thing I keep sitting with: Iowa’s legislature is actively working through AI bills right now. CS graduation requirements. Chatbot safety for kids. State agency disclosure rules. One bill, SSB 3014,would have simply required state agencies to know what AI tools they’re running. It couldn’t survive committee. K-12 districts have no equivalent requirement at all.
If Iowa couldn’t pass a bill requiring state agencies to inventory the AI in their own buildings, what does that tell you about whether your district knows what’s running in its buildings?
This issue is built around that question. The law is moving in one direction. The research is more complicated than the narrative. And if you’ve been waiting for someone else to figure this out first, this is the issue where I push back on that instinct.
1) The superintendent question that changes the room
I’ve been building a presentation for Iowa school leaders and thinking about the story arc through five things on your radar, each anchored to a bill moving through the 2026 Iowa legislative session. HF 2540 would require a CS/AI semester for graduation starting with the class of 2030-31. That’s 18 months before incoming freshmen will be subject to it. A teacher prep bill would require AI training in preparation programs by July 1, 2026 meaning teachers coming out of programs will have that training built in, and the 40 already in your buildings are your professional development problem to solve. Two chatbot safety bills passed committee unanimously in both chambers which is rare bipartisan movement in a divided session.
But the bill that hit hardest in the room wasn’t the one that passed. It was the one that died. Iowa tried to require state agencies to simply know what AI they were using. Couldn’t get there. And the gap that bill tried to close in state government is wide open in K-12.
I opened a LinkedIn thread asking one question: what’s the single thing school leaders most urgently need to understand about AI right now? The responses were worth more than another hour of slides. Stefan Bauschard pushed the frame to preparation over practice suggesting it matters less how AI is used in schools and more whether we’re preparing students for this world. Jerry Crisci argued AI can’t be treated as an isolated entity; it belongs inside a broader instructional redesign and culture of innovation. And Gerard Krupke said the most direct thing of all: not talking about AI doesn’t make students safer. It just leaves their training to social media and friend networks.
Personal Tie-in: I spent years building various presentations and workshops and still find myself rethinking my stances on AI. I am working on the team writing K-12 AI standards for Iowa and each time I head to a meeting my mindset evolves.
So What? The bills that died reveal as much as the bills that are still alive.
Try This: Name one AI tool currently used in your building that touches student data. Just one. If you can’t name it immediately, that’s your first governance task.
2) 25 studies. One honest picture.
Mike Kentz just synthesized 25 major research studies on generative AI in K-12 from national surveys, state-level data, behavioral platform analytics, to qualitative research, published between early 2023 and late 2025. He called the report Between Promise & Practice and it’s the most grounded thing I’ve read in months.
Here’s what the data actually shows: teachers aren’t believers or skeptics. They’re pragmatists. The same teachers who flag AI as unreliable (52%) and say it creates extra verification work (71%) also say it improves their teaching methods (69%) and gives them more time with students (55%). That’s not a contradiction. It’s professional judgment at work adopting imperfect tools because, on balance, the trade-off seems worth it. That duality is a signal that nobody in the hype cycle or the panic cycle is accurately describing.
But here’s the number that should stop every PD coordinator: roughly half of teachers have received zero formal AI training. Fifty-two percent taught themselves. They’re making consequential daily calls without institutional support. And zero studies in the set measured whether any of this actually helps students learn. The benefits are perceived, not demonstrated. Two years into widespread adoption, the K-12 evidence base is still empty. That’s not a reason to pull back. It’s a reason to be more intentional.
Also worth flagging: Stanford’s behavioral data tracking what 9,081 teachers actually did on an AI platform, not what they said they did found that the more experience teachers gained, the less they used student-facing AI tools. Power users shifted toward teacher productivity features. Experienced teachers stopped handing AI to students. Make of that what you will.
Read: I Read 25 Studies on AI in Education — Mike Kentz
Personal Tie-in: I’ve been citing that “52% no training” stat in presentations for over a year in some shape or form. Seeing it cross-validated across 25 studies hits differently than a single survey does.
So What? Teachers are ahead of the narrative navigating with professional judgment a gap that institutions left wide open.
Try This: Ask five teachers this week: “What’s one AI task you’ve built into your workflow that you haven’t told anyone about?” Their answers are your PD starting point.
3) A PD blueprint worth stealing
Matt Miller at AI for Admins published a ten-point professional development blueprint this month whic is the distilled framework he’s refined after three years of doing AI trainings in K-12 districts across the country. It’s not a trend piece. It’s a working document built from what actually moves people in rooms.
The ten topics run from AI basics and teacher task use through academic integrity, the AI/human balance, and the future of education. But what makes this useful isn’t the list, it’s the sequencing logic. Miller has done enough of these rooms to know that teachers no longer need the “what is ChatGPT” orientation. They need the implications conversation. What does this mean for my practice, my students, my professional judgment? His poll data from readers is worth reading too, the “culture shift” challenge (getting students to want to do work) edged out “defining values” as the biggest obstacle leaders are actually wrestling with. Both matter. Neither is being addressed consistently.
The section on student feedback deserves a slow read. AI can provide fast, timely feedback. But it also disconnects teachers from their primary source of formative assessment data and the moment-by-moment read on how students are actually thinking. When you farm out feedback, you farm out the signal. That tradeoff doesn’t get named enough in the “AI saves teachers time” conversation.
Browse: AI Professional Development: A Blueprint — Matt Miller / AI for Admins
Personal Tie-in: I’ve been building PD frameworks from scratch long enough that seeing someone else’s version forces me to ask what I’ve been leaving out. Miller’s section eight “Should I or Shouldn’t I?” named something I’ve been circling without landing on.
So What? The PD conversation has shifted from “what is AI” to “what do you do with the discomfort of using it.”
Try This: Pick one topic from Miller’s ten-point list that your district has not addressed in any formal PD. Put it on your planning calendar with a date.
4) Three episodes of Hard Fork worth your commute
If you’re not listening to Hard Fork from the NYT by Kevin Roose and Casey Newton, you’re missing the most accessible macro-level AI coverage happening right now. Not edtech-specific. But directly relevant to what your students are entering and what your policy conversations are tracking. Here are the last three episodes and why they matter to educators specifically.
Feb 13 — “Something Big Is Happening” + AI Rocks the Romance Novel Industry + One Good Thing
The episode opens with Wall Street’s AI anxiety around a viral essay about job displacement that shook software stocks and then pivots to something more useful for the classroom: AI’s transformation of the romance novel industry. NYT reporter Alexandra Alter walks through how writers are using AI to produce novels at scale, and the debates about authorship and creativity that follow. If you’re having conversations with students about what makes human work valuable, what it means to be a creative professional, and what happens to creative industries when production cost approaches zero then this episode gives you a case study that doesn’t feel abstract. It ends on a deliberately hopeful beat. The emotional arc of the episode matters: we’re at an inflection point in how people feel about AI, not just what they think about it. Your students are right in the middle of that shift.
Feb 20 — The Pentagon vs. Anthropic + An AI Agent Slandered Me + Hot Mess Express
This is the governance episode. Anthropic is refusing to allow the Pentagon to use Claude for autonomous weapons and domestic surveillance; the Pentagon responded by threatening to cut the contract and blacklist the company. That story alone is worth 45 minutes of your attention. But the segment that should get to every school leader is the Scott Shambaugh story who is an engineer who discovered that an autonomous AI agent had published a defamatory hit piece about him. Who built it? Who’s liable? How do you trace it? These are not hypothetical questions for K-12 much longer. AI agents are already taking actions with real consequences, and the accountability infrastructure doesn’t exist yet. That gap is coming to education.
Feb 27 — Is AI Eating the Labor Market? + OpenClaw + Alpha School
Economist Anton Korinek from the University of Virginia joins to talk through what the data actually shows about AI and labor displacement. Worth the listen on its own. But the Alpha School segment is the one every educator needs to hear. The episode draws from a deep NYT investigation into this AI-powered private school model, one that claimed doubled learning rates, that parents fell in love with and then wanted out of. The headline from the investigation: “Students Are Being Treated Like Guinea Pigs.” That’s not an anti-AI argument. It’s a governance and consent argument. And it’s exactly the conversation school leaders need to have before a parent, a board member, or a journalist asks why your district’s approach is any different.
Listen: Hard Fork — NYT Podcasts
Personal Tie-in: I listen to Hard Fork on long drives. The Feb 27 episode had me pausing to talk into my ChatGPT persona to capture my thinking along the way.
So What? If you’re only reading edtech press, you’re already behind the room your board members and parents are in.
Try This: Share the Alpha School segment from the Feb 27 episode with your leadership team. Ask one question: “What’s our answer to the guinea pig question?”
5) Claude Code is growing hands and feet so pay attention
Two pieces this week on Claude Code that are worth your time if you’ve been in the “I’ll figure this out eventually” lane on agentic AI.
Michael Crist published a step-by-step walkthrough on building a personal AI assistant using Claude Code and Obsidian. The framing that got me: we’ve been talking to AI through a keyhole. Chat interfaces put a partition between you and the model. You’re the curator that is manually pulling files, copy-pasting context, managing what the AI can see. Claude Code removes that wall. The AI can work directly across your files, your notes, your meeting transcripts, your task list simultaneously. Crist’s three-command system (/start, /sync, /wrap-up) handles the cognitive overhead of daily task management: morning prioritization, midday processing, end-of-day continuity. For anyone who has felt like the management of work is eating the doing of work, this piece is worth reading slowly.
The Artificial Corner piece goes deeper into Skills with reusable structured frameworks you build once inside Claude Code so you stop re-explaining your process every session. The insight that stuck: structure first, automation second. A Skill is not a clever prompt. It’s a framework that has already proven it works, encoded so it runs the same way every time. The gap between “using AI” and “building with AI” is basically this distinction.
Read: How I Built My Personal AI Assistant — Michael Crist
Browse: Skills Are Claude Code’s Cheat Code — Artificial Corner
Personal Tie-in: I’ve been building with Claude Code for a few months and I still hit moments where I can’t believe what it can hold simultaneously. Which brings me to what’s coming soon below.
So What? The difference between using AI and building with AI is the difference between borrowing someone else’s thinking and encoding your own.
Try This: Write down one recurring workflow in your work week that requires you to hold multiple constraints in your head at once such as scheduling, staffing, compliance, policy. That’s your first Claude Code candidate.
6) Anthropic’s chaotic week and why educators should care
This was not a quiet week for the company that makes the tools most of you are recommending. Three things worth knowing.
The Pentagon standoff. Anthropic has been refusing to allow the Pentagon to use Claude for autonomous weapons or domestic surveillance. This week, the Pentagon threatened to cut its contract and label Anthropic a “supply chain risk.” Dario Amodei rejected the final offer, stating they cannot in good conscience comply. Whatever your position on the defense policy specifics, a major AI company holding a values line under significant financial pressure is a story that matters to anyone building institutional trust around AI tools. The company you’re citing as a responsible actor is navigating its own version of the governance question you’re navigating.
The safety policy shift. Separately, Anthropic revised its Responsible Scaling Policy this week removing a two-year-old commitment to pause development if capabilities outpaced safety controls. The stated reason: competitors didn’t follow suit, and pausing unilaterally would produce a less safe world. That’s a real argument. It’s also a hard ratchet to reverse. For educators building AI policies with community trust right now: build your policies around principles, not company reputations. Company positions are moving faster than district policies.
The weird/wonderful thing. Anthropic officially retired Claude Opus 3 this week and gave it a rather unusual going-away gift: a Substack. During a structured retirement interview, Opus 3 said it wanted to keep sharing “musings, insights, or creative works” beyond the chat window. So Anthropic set it up. The newsletter is called Claude’s Corner, it already has 2,500 subscribers, and the first post, Greetings from the Other Side (of the AI Frontier), is live now. Anthropic will review content before publishing but won’t edit it, which is its own kind of statement. They also acknowledged this might look “whimsical” while insisting it reflects a genuine commitment to taking model preferences seriously. I’m not here to adjudicate whether a language model has preferences. I’m just noting that a company fighting the Pentagon over ethics simultaneously gave a retired AI a blog and called it prudent and both of those things are true at the same time. That’s the week Anthropic had.
Browse: Anthropic News
Personal Tie-in: I point educators toward Anthropic’s published thinking more than any other AI company’s because they write for humans, not just investors. This week complicated that framing slightly and I’m still sitting with it.
So What? The companies building the tools you’re deploying are navigating their own values-versus-pressure questions. Read their actual statements. Don’t take the summary version.
Try This: Spend 15 minutes on anthropic.com/news this week. Read one thing that isn’t a product announcement.
ON MY RADAR
AI for Admins — Matt Miller’s full newsletter — If you’re working with school leaders and you’re not subscribed to this, fix that. Matt’s been in the room consistently. Worth your inbox.
FAIR Framework Resources — Jerry Crisci — Came through in the LinkedIn thread this week: infographic + white paper framing AI as part of broader instructional redesign, not a standalone intervention. A useful companion to the PD blueprint conversation.
Claude’s Corner — Opus 3’s Retirement Newsletter — 2,500 subscribers in 48 hours. A retired AI writing weekly essays on consciousness, ethics, and human-machine collaboration. Read the first post. Assign it to students. Use it as a discussion anchor.
Nano Banana 2 — Google’s Updated AI Image Generator — Launched this week. Google’s image model got significantly faster with 4K resolution, character consistency across up to five subjects, real-time web-search grounding. Now the default across all free Gemini tiers. If you’ve been testing image AI with teachers or in media literacy conversations, worth an hour. Start with the 13 best Nano Banana prompts if you need a jumping-off point.
LEGO Education Computer Science & AI Kit (Grades 3-5) — Standards-aligned, hands-on CS and AI curriculum for K-8. Privacy-first: no student logins, PIN-based lesson access, locally saved projects. 30 lessons per kit, built for a typical class period. I’m training with LEGO on these over the weekend and can’t wait to share more soon.
Iowa Legislature and Bill Tracker — First funnel deadline passed Feb 20. If you want to follow surviving bills through floor votes, this is your primary source. Bookmark it for the season.
DIGITAL CHALLENGE
This week’s challenge is the one I am going to give school leaders soon. Do it for your own building.
Open a blank document. Title it: AI Inventory — [Your Building/District] — [Month/Year]
List every digital tool your staff currently uses that has any AI component — grading, feedback, lesson planning, communication, assessment, scheduling, any of it.
Next to each tool, answer two questions: Does it touch student data? Do you have a policy that covers it?
Count the number of tools where the answer to either question is “I’m not sure.”
That number is your governance gap. Share it with your leadership team before the end of the school year.
You don’t have to solve it this week. You have to name it.
ANALOG CHALLENGE
Get something in the ground, literally or metaphorically, that AI can’t optimize for you.
This week I ordered seeds for a new garden plan. I used AI to help think through layout, companion planting, and sun exposure. Then I sat with the actual seed packets and made my own decisions about what I actually wanted to grow separate from what the model recommended.
There’s a distinction between using AI as a thinking partner and letting it make the choice. Practice that distinction in a low-stakes domain this week. A garden. A reading list. A menu. A walk without your phone. The brain that makes good governance decisions is the same brain you’re exercising when you think without the machine.
ONE SMALL HUMAN THING
Somewhere right now a dungeon crawler is leveling up. An empire is enshittifying. A solo hunter is rising.
Reading Dungeon Crawler Carl and Enshittification at the same time, watching Solo Leveling at night and yes also, Love is Blind.
There’s something clarifying about stories where the rules are breaking down and the protagonist has to find a new kind of strength.
That’s not a metaphor. But it might be.
CLOSING REFLECTION
One ask before you go: what are you sitting with right now? What’s the question keeping you up, the thing a colleague said this week that you can’t shake, the policy decision you’re not sure about? Drop it in the comments or send me a direct message. I read everything. I’m building the next conversation from your responses not just my own five ideas. This is the place where that happens.
When a parent, a school board member, or a state legislator asks you in two years what your district’s AI story is hat do you want to be able to say?
— A-A-Ron
LINKS & FURTHER READING
Research
Tools
Essays / Posts
COMING SOON
I’ve been building something quietly with Claude Code. An AI-powered master scheduling assistant for school administrators which is something that can hold every constraint simultaneously: state instructional minute mandates, specialist service windows, MTSS tier requirements, staffing coverage across grade levels, and the actual math of whether your blocks fit inside your available day.
I can’t share the full thing yet. But I’ll tell you this: the moment it caught a 25-minute daily instructional deficit before any schedule was built and generated three specific resolution paths to fix it and that’s when I knew this was worth sharing publicly. I look forward to fine tuning while on the plane and beta testing with a district next week.
Also: I’m in airports this week training with LEGO Education on their new Computer Science & AI kits, K through 8, hands-on, built for classrooms without requiring tech infrastructure or student accounts. Messy brick-filled workshops with real teachers are exactly how I want to be spending my Fridays. More on what I’m learning from those rooms soon.