@mattbarnum.bsky.social nails it here.

This is a good structure for a piece on why AI is unlikely to replace teachers at scale any time soon. Those claims always seem to be based on naive understandings of what people actually do in their jobs. worksinprogress.co/issue/the-al...

It's hard to measure things with high school students than people realize or care to admit. We don't have evidence that levels of effort on NAEP itself are consistent over time, and we do have circumstantial evidence (attendance rates) that engagement with school overall has been declining.

Google’s New “Learn Your Way” Could Be Very Useful In Making Boring Textbooks More Accessible. larryferlazzo.edublogs.org/2025/09/16/g...

One of the best pieces of advice on teaching I ever received was to frequently ask the class, "Why would an intelligent person choose ___?"

I'm not optimistic about the odds of a national definition of AI literacy that is clear enough to actually assess. State level definitions have more promise, but I think this is something that districts likely need to be thinking about themselves. hechingerreport.org/opinion-scho...

I'm finding that using GPT 5-Thinking for office tasks is a bit like collaborating with a colleague who doesn't like to process things out loud. You ask a question or pose an idea, wait through some silence, and then receive a complete thought.

In the short term, more blue book-style assessment seems like the play in K-12. In the medium term, I'm more convinced by the ideas here.

It's a fun place to watch the range of reactions people have!

"[S]tudent expertise does not develop via osmosis or unstructured learning." I'm so glad to see that show up in Educational Leadership @ascdofficial.bsky.social. www.ascd.org/el/articles/...

"When districts say “no” to participating in research, the entire field misses out." It's hard for researchers to know which districts can support participation. (We can!) Schools Need Evidence-Based Curriculum. Researchers Need Schools to Help www.the74million.org/article/scho...

This is the first piece I've seen directly address the risks of AI on oral presentations. *Any* task that leaves the classroom can be compromised as a measure of student thinking. Good ideas here for adjusting in ways that help preserve assessment value. sarahfindlater.substack.com/p/ai-and-ora...

Highly recommended!

Anyone else have the experience of meeting with an EdTech vendor to tell them what they should start developing? “I know what you’re selling, and it’s not what we’re going to need for much longer. I think you should work on building this other thing instead that no one else is doing…”

It would be great to get this work (and @mrzachg.bsky.social specifically!) better known in the US. There are a lot of reasons that's going to be challenging, and not enough conversations *with districts* about why that is and how to move forward. www.the74million.org/article/cogn...

This is really interesting (and surprising!) work.

This centering would be a welcome change!

I'm really excited to see this work. Work around the model-implementation gap is so important for improving both healthcare and education. openai.com/index/ai-cli...

"[I]f you only read work coming from one field, you’re sometimes going to get a skewed sense of the issues and the evidence." I spend a lot of my professional time on this point. Different goals are sought in different fields, and a field's relative prestige can make a huge difference in awareness.

The point about extrapolating being a big guess is also true when folks attempt to turn effect sizes into "months of learning." When we try using dimensional analysis to make something more clear/impressive, we always lose something…

Latest from us - we've completed some more AI assessment projects and are yet to find an AI error. All the big human-AI disagreements are human error. Take a look at a couple of examples and see who you agree with... substack.nomoremarking.com/p/can-you-fi...

Really interesting paper. I think community sentiment re: collective bargaining is under-explored. People might love their teachers or school, but they rarely love district administration. When disagreements about compensation structures arise, there's a community bias in favor of current teachers.

Systematic inclusion: Is literally everyone thinking, talking, practising, learning? How much does it matter to you? teacherhead.com/2025/06/25/s...

This will be useful as we look at a potential assessment audit in 2026!

AI literacy is recognizing that successful use of genAI differs from traditional computer use. If you're looking for the "one true answer" from genAI, you're using it wrong. What Does 'AI Literacy' Really Mean and How Do We Know Students Are Gaining It? open.substack.com/pub/sarahfin...

"Is maths homework dead?" Without the changes suggested here, it sure is. I'm a big fan of #4 personally. eedi.substack.com/p/is-maths-h...

I tried making this point less articulately with our math teachers prior to their curriculum review this spring. The goals of math instruction and how to measure them aren't agreed on, which leads to wildly different recommendations about how to better achieve those ill-defined goals.

Going through some documents in my Drive and came across a report that was way too robotic - obviously from AI. It was me from June 2022.

"The more you revisit, the more you reveal." I love the posts published by Sarah Findlater. Insightful, interesting & informative. If you haven't subscribed to her substack yet I highly recommend it! sarahfindlater.substack.com/p/why-ai-in-...

"Part of the reason some people are too pessimistic is that they haven’t played around with the smartest models operating in the domains that they’re most competent in." www.dwarkesh.com/p/timelines-...

🚨New paper out! We use large language models to analyze 11,000+ pieces of preservice teacher (PST) feedback and reflections from a Texas teacher prep program. What did we learn about supervisor feedback quality? 🧵 👇 📄: edworkingpapers.com/ai25-1203 ✍️: Me, Brendan Bartanen, and @profkwok.bsky.social

"To read deeply is to insist that some things, wisdom, empathy, the expansion of human understanding, cannot be optimised, only experienced. It is to choose the difficult gift of another's mind over the easy comfort of our own."

"The state also doesn’t yet know whether the pathway, or pathways, students completed were connected to their post-high school plan, which they must create to graduate." The state's data system is not capable of distinguishing between pathway(s) a student completed and one(s) tied to their plan.

The study may have tested "recent exposure" rather than knowledge. The methodology is elegant, but the implications are limited imo. My hunch is that knowledge only really supports learning when it’s structured, consolidated, and integrated over time.

I agree with most (all?) of the critiques I've read about this study. The thing that most bothers me is when the commentary acts like these issues are somehow unique to research on AI or tech in education. So many popular education publications are similarly fishy!

🔥 15 hottest edu-research papers from the last 15 weeks: (all open source 🔓) ↓

Building on this idea, I've found interesting responses to questions about focusing an existing meta-analysis on my own context (K-12 instead of higher ed). o3 is well aware of the implications and limitations in applying these findings to K-12, for instance. link.springer.com/article/10.1...

Many of the most important “prompt engineering” skills are just management skills: clearly understanding the task to be done and what information is needed to do it; explaining the task to the AI; giving useful feedback to improve outputs; & generalizing lessons learned into a process.

I appreciate this post from @mikekentz.bsky.social about the importance of AI personality and that the differences there matter. Let's not act like all of the tools are (or should be) identical. mikekentz.substack.com/p/ai-persona...

I use AI for all kinds of things, but the most common is likely: "Make a pivot table showing median values." Not something that's going to show up as a big difference for the organization as a whole, but simplifying a tedious task feels like a meaningful improvement at an individual level.

As part of a larger standards-based grading movement, these practices assume curriculum, instruction, and assessments that are all tightly aligned to established standards. That's rarely (never?) the case across an entire institution and considerably more complicated than proponents admit.

We shouldn't wonder why young people are leaning so heavily on AI for school assignments when corporate leaders are telling young people--what you can do isn't as good as what I can get from AI.

I appreciate @pepsmccrea.bsky.social putting out such a great post on my birthday. Whether AI can act as a decent tutor is not an interesting or valuable question to answer. How it can help with learning (professional or otherwise) is far more interesting. snacks.pepsmccrea.com/p/motivation...

It would be nice if these panels included more people actually working in school districts. Even people who worked in one at some point in the recent past…

"Be authentic" always struck me as strange advice for a teacher. In my mind, "Mr. Cantonwine" was a character played by a guy named James. To my knowledge, students were not aware of the differences between us.

Well-being and technology: The effect of individual laptops in high school www.sciencedirect.com/science/arti...

I appreciate @mrrablin.bsky.social's point here that success (i.e. an early win) precedes motivation more than the other way around. www.innerdrive.co.uk/blog/about-s...

It would be really interesting to see a similar study looking at principal and superintendent candidates.

The fourth bullet has been a big one so far. Teachers have (understandably) wanted to know what AI could do *for them* before leading students in using AI. Interestingly, focusing on AI as a way to support teacher wellness has increased interest from some otherwise tech-resistant staff members.