AI Is Everything Everywhere All At Once

A Case for a Different Mindset

One thing is clear - AI cannot shovel snow.

Agentic AI is currently all the rage - coding tools released by OpenAI and Anthropic over the past six weeks have driven the conversation across legacy media and Substack alike. It does not require accepting the hype to acknowledge that something has shifted. As I sit and watch a late February blizzard blanket my neighborhood, I’ve been reading and thinking about what the rise of agentic AI means for teachers, students, and schools, and it underscores something I’ve been reluctant to say outright.

Practically since its arrival, schools have treated AI mostly as a problem - how do we stop students from using it to cheat? What can we do to slow all this down? How can schools design policies to prevent AI from destroying authentic student work? These are real and ongoing issues and they deserve serious treatment. I’ve catalogued many of them.¹ But the “AI is all negative for schools” stance has hardened into orthodoxy and I worry that it’s making it harder to see anything else.

The increasing recognition that AI is evolving from a series of chat interactions into the ability to build entire projects from scratch introduces a fundamentally different question. And that question is going to require more imagination than schools may be capable of given where most teachers currently are.

For starters, I simply don’t think that many educators are aware of how quickly the landscape is shifting. I applaud those teachers who have paid enough attention to try to get more knowledgeable and fluent with the more advanced models. But it seems that just when many schools are finally deciding it’s time to confront the “AI issue” - either by paying for AI wrappers like Magic School (a decision with more consequences than many realize) or bringing in outside consultants to review the basics - the conversation is moving elsewhere. It doesn’t mean that learning the basics and understanding the strengths and weaknesses of LLMs isn’t important. It’s just that, for anyone who follows this closely, it’s clear it isn’t going to be enough.²

I’ve watched the response to agentic AI unfold over the past several weeks across a dozen Substacks and podcasts, and I’m struck by how much of it has followed a familiar pattern: here is a powerful new capability, but the focus in schools is mostly on why it’s dangerous. The concerns raised - more cheating, student data privacy, and the security risks of connecting AI agents to institutional accounts - are all legitimate and will need to be addressed. But when the primary response to a genuinely new development is to mostly fixate on the ways it might go wrong, I think the conversation is incomplete. A preoccupation with the downsides that crowds out everything else may prevent educators from recognizing the potential usefulness of a transformative technology.

Seeing Is Believing

I remember the first time I saw a student share a Google Slides presentation from the cloud. It seemed like magic at the time - the ability to host digital work that could be accessed anytime, anywhere, from any machine. I hadn’t used the Google suite of products very much in the early aughts and that one demonstration convinced me that cloud computing was the future. I needed to see it to understand it. Google Drive was once unknown to me, as it was to most teachers, but it’s now an essential part of how I store, organize, and work with all my digital materials. It happened over many years and now I can’t imagine living without it.

Technology works like that. You can read as much as you want, but until you see what everyone is talking about, until you watch an AI agent design and deploy an online module or website in real time, until you begin to realize what this means for simply the manner in which we produce documents - spreadsheets, presentations, memos, reports, and virtually anything else that requires organizing information into precise digital formats - you are not likely to appreciate how this will affect our work habits.

What Agentic AI Has Allowed Me To Do

In my Government and Politics class, I designed an Ethics Bowl case built collaboratively with AI. I could not find a suitable article summarizing the past year of U.S. foreign policy that was not written from a hyper-partisan perspective. Instead, I co-created a thorough analysis of the current administration’s foreign policy decisions, framed through the lens of Thucydides’ Melian Dialogue and Kantian ethics. The case went through multiple drafts across several days. I subsequently used a separate AI project to fact-check every quote, date, and statistic, catching one major error that I corrected. I am confident I could not have produced a case of that depth and rigor in the time I had without AI collaboration.

I also wanted students to have an interactive preparation tool - something they could use independently to click through each policy area, toggle between factual summaries and competing arguments, review key terms from the textbook chapter, and test their comprehension before the live discussion. Six months ago it would have been impossible. Building an interactive web application is not something I knew how to do. I don’t know how to code.

Using Anthropic’s Cowork - one of the agentic tools that everyone is currently writing about - I described what I wanted. I uploaded my case document and assignment instructions. And over the course of a few working sessions, the AI built it. A functioning interactive web application with tabbed navigation, expandable case sections, a glossary connecting textbook vocabulary to the specific case, verified source citations, and a comprehension quiz with immediate feedback. I had it hosted at a live URL and shared it with students through our learning management system.³

The same session also produced a fifteen-slide introductory presentation for the activity. All from the same source documents.

The process was not seamless and I needed to troubleshoot a handful of steps. The tool needed my professional judgment - what to include, what to cut, when something wasn’t right, how to redirect. It took some time. But it was work that produced something I could not have produced alone, and it was work that would have been unthinkable for a classroom teacher even a year ago.

That was the moment when the possibilities crystallized.

The Gap

Many teachers don’t have the time or inclination to do this and many won’t for months or even years.⁴

I’ve had a curiosity about digital tools from an early age and invested in the conditions to make it possible to create the use case described above: a paid subscription to advanced frontier models, custom environments I’ve built over months of experimentation, and a specific pedagogical need that gave me a reason to investigate what the tools can do right now. I’ve spoken with enough colleagues and led enough conversations to know that most educators haven’t encountered these tools, let alone used them.

Knowing that I have the ability to build interactive web applications means I can’t unknow it. More use cases and possibilities will keep coming. The distance between those two realities - between where most educators are right now and where the technology already allows motivated practitioners to go - is likely to create more of a divide in schools, especially if students begin to incorporate these skills into their work on their own.

Imagination As Well As Caution

The dominant response to AI in the education space - currently extending to agentic AI - is primarily to focus on what’s gone wrong.

And I understand why agentic AI is so scary for schools - when someone describes a tool that can autonomously access a student’s Google Drive, browse the web, and execute tasks without supervision, the immediate instinct is to worry about data privacy and security. That makes sense.

However, I’d love to see the conversation broaden. What I’m arguing for is imagination as well as caution. Believe me, I understand that generative AI, specifically agentic AI, represents significantly more threats than Google Drive. But if we want to treat it as “normal technology,” as many who resist the hype are preaching, then we need to understand the typical progression that “normal technology” takes. I’m not suggesting we ignore the risks. I’m suggesting we match them with equal imagination about the possibilities.

An excessive focus on what could go wrong produces a profession that is perpetually reactive - always responding to what the technology does, never shaping what it could do. And it produces a widening gap between what educators currently think AI is (simply a cheating tool, a security risk, a privacy nightmare) and what it may actually become (a genuinely powerful instrument for creating things that were previously impossible for non-technical professionals to build). I realize this may be a counter-intuitive take at the moment, but I think it needs to be part of the dialogue. There are still people who see AI as nothing but a threat to education, but my experience and that of a growing number of practitioners suggests otherwise.

What if the question encompasses more than just “how do we protect students from agentic AI” but also “what could we eventually ask students to build with these tools that they couldn’t build before?” What if assignments evolved beyond “write an essay” or “make a presentation” toward “design an interactive tool that helps someone understand this problem?” That question requires imagination and a sense of possibility, both of which are sorely needed right now.

Where This Leaves Me

Schools are likely to escape the full implications from agentic AI, at least on the student side, for at least another year or so, if the pattern of adoption rates is anything like it’s been up until now - mostly slow, uneven, and imperfect. Teachers have not been incentivized or trained to integrate agentic workflows into their routines. But just like Google Drive, I suspect agentic AI tools will simply become part of how many of us work in the near future. I’ve already seen too much value to pretend otherwise.

Moving through the current moment as an educator is going to require more imagination than has been demonstrated up until now. The same concerns about cheating and authenticity will persist, but I suggest that instead of panic and surrender, we embrace both the uncertainty and the possibility. What AI is now capable of is genuinely helpful and, to me, remarkable - while at the same time unsettling and, for many, frightening. Those things will have to coexist.

I’ve been reluctant to make predictions in this newsletter - the field moves too fast and most of what’s coming is outside my area of expertise. But I do know the classroom and I’ve been creating curricular materials for three decades. The nature of how AI tools are evolving represents the single most interesting and exciting development on the teacher side I’ve ever experienced.

With respect to AI’s overall impact and whether it does what its most ardent proponents promise or its most informed critics predict, I still don’t know. Frankly, no one else does either.

The provocative title of this piece is a movie reference, but I also think it will be the literal truth. AI is going to be everything, everywhere, all at once. The question for educators is what we are going to do about it and how we frame it - as an ongoing problem to manage or as something that might actually change what’s possible in a classroom. I’m holding out that we can rise to the occasion. I don’t see that we have much of a choice.

---

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and organizations navigating AI integration. Take a look at my website and reach out - I’d love to hear what you’re working on.

1. The recent Brookings Institution report on AI in K-12 is the most comprehensive survey of these concerns to date. Notably, its conclusion was not that AI offers no benefits to education but that, in the current moment, the negatives outweigh the positives - a finding based on research conducted prior to the emergence of the agentic tools discussed in this piece. I wrote about the report in a previous post.

2. Ethan Mollick’s recent guide, “A Guide to Which AI to Use in the Agentic Era,” provides an excellent overview of where these tools stand right now and is worth reading for anyone trying to keep up with this space. It also demonstrates how complicated the technology is getting.

3. The case itself, “The Iron Laws of the World,” and an interactive student preparation tool built with Anthropic’s Cowork are both available online. The case itself can be found here. The interactive tool is here. I used Mike Caulfield’s SIFT verification methodology via a dedicated AI project to fact-check all quotes, dates, and statistics.

4. Some have. Historian Benjamin Breen has made far more interesting and advanced online activities than I have, including this cool history simulator.