Notes on curriculum, AI, and teaching, as I post them.

🕵️ I have started writing a little worse on purpose.

I doubt I am the only one, and I have spent a few weeks trying to work out whether what I am watching is a small stylistic tic or the front edge of something linguists will find obvious in thirty years, which is the kind of question I cannot leave alone. What I think is happening is this. We are not simply abandoning the words and punctuation that AI overuses. We are also adopting the ones it does not.

Two moves → one instinct.

The retreat is the easy half to see. I wrote a while back about deleting my own em dashes, not because they were wrong but because I did not want anyone assuming a robot had written my sentences, and the same quiet exit is now underway for "delve," for "moreover," for the tidy three-item list that used to be perfectly ordinary and now reads like a tell. None of those are errors. They are good English being dropped for social reasons, which is a thing that happens to languages, though usually across generations, not eighteen months.

The second half is the one I do not see anyone naming. The contraction is having a moment. So is the fragment. So is the typo somebody noticed and left in anyway, and the sentence that runs on too long because the writer got excited, and the aside that goes nowhere and stays in.

✍️ We are not only fleeing the machine's fingerprints. We are manufacturing our own.

The pressure all of this pushes against has been measured, which moves it out of vibes and into something I can hand a skeptic. Kobak and colleagues went through 15 million biomedical abstracts for Science Advances last year and found an abrupt spike in AI style words after 2022, estimating at least 13.5 percent of 2024 abstracts had passed through a language model, reaching 40 percent in some fields. Their comparison point is what stopped me. The effect on scientific vocabulary was larger than COVID's.

I am a historian before I am anything else, so this is the part I keep circling. The Gileadites in Judges 12 sorted Ephraimites out of a crowd by how they pronounced a single word, and armies and resistance movements have run some version of that play ever since, which suggests a language always ends up carrying a password whether or not anybody designed one.

📜 Every shibboleth on record separated humans from other humans. This is the first one built to separate us from a machine.

I have no idea what that does to English over fifty years. I know it is not nothing. I know we are all doing it without having agreed to any of it. 🦊

So tell me what you have started avoiding. I am collecting them.

🤖 I say please and thank you to AI, and I am not planning to stop.

That is a strange thing to have an opinion about, and yet here we are. There is a real argument happening right now, in staff rooms and group chats and my own feed, over whether courtesy toward a language model is a harmless habit or a small act of self-delusion. The case against it is sound. The thing has no feelings to hurt, politeness costs tokens, and manners aimed at a machine are manners spent on nothing.

Here is my rub: that case is measuring the wrong thing. My reasons have almost nothing to do with the AI.

🗣️ The first is practice. We are all talking to systems more and to people less, and speech is a skill like any other. Skills that go unused get worse. I do not want to spend four hours a day issuing flat commands and then walk into a meeting expecting my tone to remember how to be a person.

Courtesy is a muscle. You are either using it or you are not.

🔬 The second is the part I actually have evidence for. Aitken and colleagues (2026) published a meta-analysis in the Journal of Organizational Behavior pulling together 123 samples, 14,717 people, and more than 127,000 observations of how mood relates to job performance inside a single workday. Positive affect tracked with better task performance and more helping behavior. Negative affect tracked the other way.

The honest caveat is that the moment-to-moment effects were small, and the between-person effects were larger, meaning your habitual mood matters more than any single bad minute. Which is exactly my point.

Nobody is ruined by snapping at a chatbot once. Spending an entire workday in a low-grade snarl is a different animal.

🌱 The third is the one I cannot prove. I am saying so on purpose. We are training these systems on human interaction, and we do not fully understand what that training is building. Every educator I know accepts that how you speak to a learner shapes what that learner becomes. I am not claiming a model is a child. I am saying we have not earned the confidence to assume the principle stops at the keyboard, and I would rather be wrong in the generous direction.

Curriculum design taught me that tone is instructional whether or not you meant it to be. It is doing work in the room either way. 🦊

Where do you land on this one? I would love to hear it, especially from people who think I am wrong.

⚰️ Death of the Em Dash

The em dash used to be one of my favorite tools, the little pause that lets a sentence pivot, add a thought, or land a punchline without starting over. Then ChatGPT and Claude fell in love with it too, and now it is basically their signature move. Which is honestly a little funny, because it means one of the most useful pieces of punctuation in English got typecast as "the AI one" almost overnight.

This is a post about a punctuation mark having a full identity crisis. 🥸

Here is my rub: I was doing this way before it was cool, or suspicious. Educator, grad student, professional over-explainer, my brain has always pivoted mid-sentence like it forgot where it was going. Now I catch myself deleting my own dashes, not because they are wrong, but because I do not want anyone assuming a robot wrote my sentences.

🤖 That's a genuinely strange thing about writing in 2026. It is not that AI writes badly, the opposite, in fact. It is that un-customized AI writes so recognizably that it changed how the rest of us have to write to still sound like ourselves.

✍️ So this is less "AI ruined punctuation" and more "I miss my dash, and also, kind of respect that it has main character energy right now." If you have caught yourself editing your own natural style for the same reason, I would love to hear it!

📊 New landscape scan from FullScale looked at 106 high schools across 30 states doing competency-based learning. Their core finding: "If we cannot clearly define and see competency-based learning in practice, we cannot effectively support, measure, or scale it."

That's the same problem my dissertation is asking, just at a different scale.

I'm investigating why the same AP course, taught by the same instructor, with the same resources, produces inconsistent exam outcomes across years. This report is asking why competency-based learning produces inconsistent outcomes across schools. Same structure, different zoom level: you cannot fix what you cannot see, and most systems are built to see the easy parts, not the parts that actually determine quality.

The finding that stuck with me most: schools describe their goals and design philosophy for CBL far more often than they describe the actual systems that make it work, assessment, mastery progression, aligned supports. That gap is exactly where quality lives or dies. Anyone can write a mission statement about mastery. Far fewer can show you the rubric that actually measures it.

📚 It also validates something I have believed for a while working inside AP assessment design: multiple choice alone was never going to capture mastery. AP's answer is a messier, harder-to-scale mix of FRQs, DBQs, LEQs, and rubric-scored performance tasks, because that is what it actually takes to see whether a student can do the thing, not just recognize the thing. CBL is running into the same wall at a much bigger scale, across 30 states, with far less shared infrastructure to solve it.

🔬 If the field wants CBL to be more than a mission statement, the unglamorous work is assessment design and progression systems, not just redesigning the student experience.

Worth a full read if you work anywhere near high school redesign, assessment, or ed policy.

🎭 I grew up in theater, and I never really recovered.

Translation → zero shame, a permanent instinct to hold a room, and an internal volume knob that only goes to "projected to the back row."

Extremely useful in a classroom. A genuine liability in a library. 📚🙅‍♀️

I teach history like a spill-the-tea session. Less "here are the causes of the war" → more full reality-TV-recap energy, gasps included. Same content. Completely different level of attention in the room.

Ten years of that taught me the one rule that survives every curriculum framework I've ever read, most of which were clearly written by people who never had to compete with a phone for a teenager's attention span:

🎯 If it does not hold attention, it does not teach anything.

Curriculum design gets treated like a compliance exercise. Check the standard ✅ write the objective ✅ laminate something, go home. The actual job is getting inside how a mind learns something and figuring out the sequence that makes it stick, whether that's a 16-year-old in AP History or an adult sitting through a professional development session they would very much rather not be at.

I've built a UC-approved course from a blank page. I ran two years of formal action research on an AI tool before I trusted it near my students, because trust me, it's good is not a methodology. I write curriculum for subjects that were never mine to teach because someone needed it done and I am, apparently, constitutionally incapable of saying no.

None of that is separate from the theater kid thing. Same instinct → different stage → minus the jazz hands. Mostly. 🦊

I wish artificial intelligence did not exist. Students need to be taught to use it anyway.

That's not the sentence people expect from someone who talks this much about AI in education, but it's the honest starting point.

For most of human history, understanding something hard required sitting with it. Working it out. Getting it wrong, trying again, and building the kind of reasoning that only comes from struggle. AI has made that struggle optional. Ask a question, get an answer, move on. I watch what that instant access is doing to critical thinking, and it worries me.

🧠 A mind that never has to wonder is a mind that is losing practice at wondering.

That's one half of what I believe. Here's the other half.

I'm a historian before I'm anything else, and history has taught me that resisting the direction something is moving in is rarely the winning strategy. Technologies that fundamentally reorganize how people think don't ask permission first. AI isn't going away, and it isn't slowing down. It's getting stronger, spreading faster, and operating with remarkably few guardrails around it, which is its own kind of dangerous.

Wishing it away isn't a strategy. Understanding it well enough to shape how the next generation uses it is.

So yes, I believe AI belongs in the classroom. Not as a replacement for thinking, but as something students are taught to use deliberately, the way you'd teach someone to use any powerful tool: with a clear sense of what it's good for, what it costs when it's overused, and where the line sits between using it and outsourcing your own mind to it.

Students need to leave school knowing how to use AI to benefit the people around them, not just to finish homework faster.

📚 That distinction is going to define this generation the way literacy defined the last several.

What I keep coming back to is that we're not just watching a new tool enter the classroom. We're standing at a genuine precipice, one where AI is going to change how people process information at a cognitive level we don't fully understand yet. That's not hyperbole. That's a research problem, and an enormous one.

🔬 Educators are going to need new studies, new frameworks, and a willingness to be wrong and adjust as we learn what actually happens to a developing brain that grows up with an answer engine in its pocket.

I don't think the answer is optimism or alarm. I think it's rigor.

History doesn't wait for anyone to be ready for it, and neither will this. I'd rather be one of the people figuring out how to move forward with it than one of the people still arguing about whether we should.