When Infrastructure Becomes Destiny

How Schools are Losing the Battle to Stay Ahead of the AI Curve

Created with Nano Banana Pro

Over the past week, two substack essays brought home what I’ve been sensing for months. The first covered Google’s AI expansion into educational infrastructure through Workspace for Education, Guided Learning, and Google Classroom, all integrating its newest model, Gemini 3. The second piece, published the next day, documents OpenAI’s release of ChatGPT for Teachers. I’m grateful to both writers for covering these developments in such detail. What follows is my attempt to synthesize what these moves mean for schools facing an AI onslaught they can’t realistically control.

A lively debate among educators is whether AI’s adoption in schools is “inevitable.” Marc Watkins argued persuasively in January that AI is “unavoidable but not inevitable.” This past week made me question whether that distinction really matters anymore. The aggressive push by both Google and OpenAI into education - both K-12 and higher ed - is happening so quickly that many institutions barely have time to make informed choices before being backed into a corner to make decisions that have longer term implications than they imagine.

Recent product launches by both Google and OpenAI, documented meticulously by Wess Trabelsi and Jacob Carr, make clear just what’s happening.

The Models Are Still Improving

On November 18th, Google released Gemini 3. Just a few months ago, all eyes were on GPT-5 - until it landed with a thud. Dismissed as underwhelming, GPT-5 seemed to confirm fears that AI progress was slowing.

Gemini 3’s reception has gotten all the attention GPT-5 missed, setting a new bar for the frontier models.¹ Gemini’s upgraded image generation tools - faithful infographics, accurate diagrams, and integration with video - are also being hailed as significant breakthroughs.

Skeptics saw GPT-5’s underperformance as proof AI was plateauing, but they missed the bigger picture: Gemini quietly leapfrogged it. The competitive cycle ensures continuous model improvements among multiple companies desperate to stay on top, creating a pace that the public - let alone educators - simply cannot keep up with.²

Why does this matter? On an individual model standpoint, it doesn’t. Most students aren’t aware of the newest models and most teachers won’t be either. But it means that where we will be landing at the end of this school year versus the start will continue to put distance between those educators who’ve engaged seriously with AI and those who haven’t.

For schools hoping AI progress would slow down and give them time to deliberate, Gemini 3 ended that possibility. The challenges raised by corporate product cycles that move in weeks compared with institutional decision-making that operates in school years makes it virtually impossible for them to make informed decisions.

The Google suite of products is already embedded in almost everything teachers do. Though many teachers are aware of Gemini, its market share remains minimal compared with ChatGPT’s dominance.³ Google hopes its newest flagship model will change that over the next few months.

Two Announcements, One Strategy

But the real significance of this past week isn’t Gemini 3’s improved technical capabilities. It’s the implications of these two companies’ strategic outlook.

The day after Gemini 3 launched, OpenAI announced ChatGPT for Teachers. They explicitly positioned it as “the safe AI for educators.” Jacob Carr’s piece underscores its significance:

ChatGPT for Teachers is not just free access. It is a walled-off, educator-only version of ChatGPT built on top of the 5.1 model with a dedicated reasoning model, unlimited custom GPTs, connectors to Google Drive and Microsoft 365, and clear data-use protections. It gives teachers the full power of ChatGPT inside an environment where student data is protected, educator content cannot be used for model training, and districts can add oversight when they choose.

ChatGPT for Teachers Just Launched

OpenAI isn’t just offering a tool for teachers. They’re claiming to solve a problem they themselves created when ChatGPT launched in November 2022 with no warning, no guardrails, or guidance for schools. Last year they offered their best model free to students. Now they’re doing the same for teachers, making it almost impossible for schools facing budget shortages not to consider adopting a free, high quality AI model for their staff, even if many are still ambivalent on whether or how to use it.

Within the same week, Google made its own coordinated move, as Wess Trabelsi describes, as the powerful Gemini 3 becomes the default model within Google Workspace for Education and Google Classroom, both platforms many schools depend on. While Wess focuses on the implications for edtech startup companies, a larger story is the full-on trench warfare between two tech giants competing for the billion-dollar education market - with schools caught in the crossfire.

Generated with Nano Banana Pro

Encirclement and Attrition 2.0

Earlier in the summer, I borrowed the term “Encirclement and Attrition” from the NAACP’s civil rights strategy to describe OpenAI’s push into education to make ChatGPT the most trusted source for student learning. I wrote that analysis focusing on OpenAI’s Study Mode rollout which some welcomed as a salve for educators. I was more cynical and saw the writing on the wall that the big companies were mostly competing for student eyeballs as a long term profit source.

What I didn’t fully anticipate was that Study Mode may have simply been OpenAI’s initial move to disrupt Google’s potential domination of the education market. Make no mistake - both OpenAI and Google are going head to head to capture what we now know is the largest segment of ChatGPT’s user base: students. Each has made their latest model free for undergraduates over the past several academic semesters.

Both OpenAI and Google are operating under the assumption that AI will be a learning tool used by students. They are not wrong. It already is by millions.

The major question these companies care about is that it’s their AI model students are using. The core problem is this prioritizes their vision of what that learning looks like - not educators’.

This pits school decision makers between two companies racing to establish their AI as the default model before educators have finished even debating whether to use AI at all.

While OpenAI has the larger user base, Google has significant advantages: they have deeper pockets and, crucially, they’re already embedded in school infrastructure. As Wess documents, Google now offers AI-powered features in Google Workspace for Education (email, docs, sheets, slides), Gemini integration directly into Google Classroom, AI Mode as the default search experience, and Guided Learning features throughout their educational products - all powered by Gemini 3 and all on the free tier.⁴

The Google Classroom integration is especially significant because it competes directly with the AI wrappers so many schools have already adopted. Many teachers still prefer Google Classroom over their own institution’s LMS (Learning Management System) - this undercuts attempts by schools to train faculty and students on yet another platform to which some have already committed. Why bother when your entire community is already familiar with the free version?

The Timeline Problem Nobody Can Solve

Here’s where the collision between corporate and educational timelines becomes impossible to ignore. The AI models (like Gemini 3), features (like ChatGPT for Teachers) and integration (in known platforms like Google Classroom), are arriving at such dizzying speeds that school leaders, even if they were inclined to stay on top of the latest developments, simply don’t have time to analyze every choice thoroughly.

More significantly, most schools start planning for the next school year early in the academic calendar - decisions about which platforms to use or adopt must be made months in advance, something completely at odds with the way the tech companies operate. Once schools commit to one company’s product, the AI landscape will have radically changed.

Should they take advantage of the free ChatGPT for Teachers offer while at the same time train teachers on the new features coming from Google?⁵ How do schools coordinate an information environment when so many different pieces of the puzzle are coming at them at warp speed from multiple directions?

Teachers push for one tool, directors of technology have other ideas, and the decision makers at the top don’t have the expertise to offer meaningful counsel. Throw in budget considerations, state-wide and district policy requirements and alignment, and other ancillary factors, and you can see why paralysis has set in.

School planning cycles take place from one year to the next. New AI releases happen within weeks. By the time a district completes its planning for fall 2026, the competitive marketplace almost guarantees those choices will be obsolete before they’ve even been implemented.

When Infrastructure Becomes Destiny

To avoid AI integration entirely would mean replacing Google Workspace and Google Classroom, banning ChatGPT outright, retraining staff, and re-platforming - all while still teaching. It’s not going to happen.

Schools may think they’re making routine software decisions by updating productivity tools and adopting free resources for budget-constrained teachers. But these updates carry pedagogical implications that won’t become apparent until they’re embedded in daily practice. What looks like choosing between Google and Microsoft email systems or ChatGPT for Teachers is actually choosing the default AI learning environment for the foreseeable future. By the time districts realize they’ve made strategic choices about which AI platform to use, those choices will feel irreversible because they’re woven into the digital fabric that makes schools function.

All of this means the question of whether to use AI has already been answered through infrastructure dependencies and default community behavior. Schools can still choose their pedagogical response, their assessment strategies, and their values around learning, but the question of whether AI will be integrated into a school’s digital environment? I don’t think that’s up for debate anymore. Silicon Valley has made sure of that.

The Faculty-Student Fluency Gap Gets Worse

In “GPT-5 Is Coming,” I documented the growing fluency gap between students and faculty. That gap will keep widening - not necessarily because teachers are resistant (though many still are), but because we’re all getting lapped by new products no one can keep up with. The last two weeks alone have seen a blitz of significant new releases.

Even educators who experimented with ChatGPT in the past year or two are still several model generations behind. They likely haven’t used Claude Opus 4.5, Gemini 3, or ChatGPT 5.1, or any of the multimodal capabilities these models now offer. As I wrote in July, nearly 30% of educators have never used a single chatbot. If that number remains static, it will become more and more difficult to catch up.

Students are ideal early adopters because they face constant demands to produce content on unfamiliar topics, under time pressure, for stakes that matter but aren’t career-ending. They have dozens of use cases every week - explaining concepts they’re trying to learn, analyzing texts they’re reading for the first time, and generating ideas for essays on topics assigned the day before. The combination of free access, high volume, diverse subject matter, and low odds of getting caught make experimentation with AI feel low-risk. Using AI is now just a normal part of the student experience.

Teachers, by contrast, work within established areas of expertise and settled routines. They have fewer natural pressure points driving them to experiment with AI. This structural asymmetry helps explain why students are pulling ahead so quickly.

As the distance between where faculty understanding sits and where student use has reached continues to grow, it places even more pressure on schools to do something, even if that something means catering to the dubious promises of AI companies by adopting their tools. OpenAI and Google are marketing to students directly, with platforms and capabilities that match precisely what they are being asked to do, forcing the issue before teachers even weigh in.

With the release of Gemini 3, students can easily create sophisticated reports, audio overviews, slide shows, infographics, and apps for virtually anything they need. And Google is free. This is on top of ChatGPT which most are already familiar with. Many teachers, meanwhile, are still trying to make up their minds whether AI is even ethical to use at all.

What Remains Now

Marc Watkins is right that AI adoption isn’t inevitable in the sense that it needs to appear in classrooms. Teachers, at least for the moment, are still in charge of their pedagogical and assessment strategies. Most can set their own rules around AI.

But when the practical alternative to school-wide “adoption” would require abandoning core infrastructure, retraining entire faculties, blocking tools students access anyway, and rolling the boulder up the hill against the accelerated timelines set by billion-dollar companies, the difference between “inevitable” and “unavoidable” becomes meaningless. In most schools, AI appears in everyday tools whether teachers like it or not.

I’m not arguing that schools should surrender to whatever Google and OpenAI decide. Not by a long shot. But I am sounding the alarm that the window for proactive choice is rapidly closing if it hasn’t slammed shut already. Decision makers need to understand they are caught between corporate interests that are executing the famous Steve Jobs maxim: “people don’t know what they want until you show them” and their their own commitment to an educational mission that serves student learning, not corporate revenue streams.

The decisions that still matter are no longer about whether AI will be present in schools, but about how it’s going to be deployed. At least three key questions remain:

Three Key Questions

How do schools maintain agency when the tools they depend on have AI embedded by default?

This isn’t a hypothetical. Google Classroom isn’t asking schools if they want AI - it’s already arrived. Now schools must shape how it’s used rather than simply accepting whatever defaults the platform provides.

What happens as teachers continue to fall behind students while institutions deliberate?

As far as I can tell, the fluency gap isn’t closing. Every week of institutional paralysis is another week students get more sophisticated with tools their teachers don’t understand. At some point, this gap becomes unbridgeable without massive investment in professional development that most schools can’t afford or aren’t willing to invest in.

How can schools make intentional and informed choices before those choices are made for them through infrastructure defaults?

The timeline collision isn’t theoretical. Schools operating on yearly planning cycles cannot keep up with companies releasing updates weekly. The window for deliberate choice exists, but it’s measured in months, not years.

The past few weeks have made one thing crystal clear: OpenAI and Google aren’t waiting for educators to make up their minds. The integration is happening now, constantly, through infrastructure most schools can’t abandon and under time constraints they can’t match. What matters is whether schools can answer these questions intentionally - or whether infrastructure will answer them by default.

---

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. Ethan Mollick covers the ins and outs of Gemini 3 in his newest post, Three Years from GPT-3 to Gemini 3.

2. While working on this piece, Claude released Opus 4.5 on November 24th, only reinforcing the point about the cycle of competition among the top AI companies and suggesting that scaling LLM’s, though they may not lead to AGI, are nevertheless going to continue to improve for the foreseeable future.

3. According to its own “AI Overview,” its market share was just 2.2 percent compared with ChatGPT’s 82.7%.

4. It’s worth reading the entire piece, Google on the Offensive: Is This the End of AI EdTech?, for the full picture.

5. I’ve reported previously that Miami went all in on Google in spring. If they have buyer’s remorse given OpenAI’s release of ChatGPT for Teachers, it would be difficult - and costly - to make the switch.