Welcome to Base Classes
How education is changing in the age of AI
The bots are here.
AI use has exploded in education at all levels - I’m launching a newsletter to keep track of the changes.
Most teachers are already using AI
The majority of teachers are already utilizing LLMs in some form, and more each day. In June 2025, over 60% of teachers surveyed said they used AI in some form during the previous school year. By October 2025, a separate survey said that over 85% of teachers reported using AI.
Dan Carroll, co-founder of Clever, noted recently that “2025 is the year where I feel like AI went across the chasm from early adopters to something that most teachers have access to and are using, if not every day, maybe every week or every month.”
At the same time, education is a major use case for the large LLM vendors. Over the summer, Google announced that Gemini was free for teachers, and in November, OpenAI followed suit with free ChatGPT for teachers. By making it free, the products are seeing even more adoption. This makes sense: more than 10% of overall ChatGPT use is classified as Tutoring or Teaching.
And major AI-focused educational products continue to grow quickly - for instance, MagicSchool claims use by 6 million teachers (up from 5 million a few months ago).
Educational research is necessary, if a bit messy
I have been craving a deeper look at understanding how the education system is changing today, and what we are learning about it as a society that will inform where it is going in the near future.
So, last May, I started a podcast, The AI Who Taught Me, to follow all the rapid changes as LLMs gained adoption in education. We are effectively launching a huge experiment as society: what happens as we can rent intelligence for effectively pennies or free? How does that change the practice of teaching and learning?
The research world is deluged by studies about the use of AI. There are new papers released every day from all over the world, and I read as many as I can. Last year, I covered a few bombshell reports that gained significant media attention - such as the MIT study that claimed ChatGPT rots your brain, or the Nature meta-analysis about the positive impact of ChatGPT on learning. But there are so many more that fly under the radar.
Deciphering the high quality education research can be a challenge. We have to contend with publication bias, vendor-funded studies, and the fact that research can move slowly compared with the pace of change of underlying technology. Some studies released even just in the last year still reference gpt-3.5 — how can you draw conclusions about the capabilities of the latest models when using three-year-old models?
Does this mean that the academic research world is not worth paying attention to? No, because the alternative - ignoring the academic research - is even worse. Without peer review and the scientific method, we end up with vendor reports and unsubstantiated claims from whichever school or product has the largest PR team. For example, if you search “AI in education” one of the top results is Alpha School, which claims that students learn 2x their peers. However, when I dove into the results, I found that their entire “white paper” is a sham - the results come from using the wrong column in their report. At least with peer-reviewed papers we have a shot at discerning what is really working, even though academic research may be a messy process.
Yet, I’ve found that podcast is not the best medium for in-depth, detailed discussion of research. So I am starting a newsletter to go deeper.
Welcome to Base Classes
The name, Base Classes, has a double meaning. In programming, the base class is the template from which instances of an object can be copied. A base class provides the core principles that define the rest of the whole system.
In education, base classes are be the core ways of operating, many of which have been consistent for decades or longer: students are grouped in classes with peers of the same age, learn from one or two teachers in classes of 20 to 40 students, the day is oriented around time blocks, subjects siloed and departmentalized, on static curriculum pathways, with periodic high stakes assessments.
With each wave of technological and societal change, the education system adapts. Last decade, we saw enormous adoption of laptops and phones, and an explosion of edtech companies that offer more products than were available in the past. In classrooms today, students routinely use laptops to access robust educational programs, pull from a much wider variety of publishers and resources than they had access to 20 years ago, sometimes have automatic differentiation, and assessments are now largely taken online. These changes have had a huge impact in classrooms.
But when we have the rise of fully automated intelligence, will those primitives still hold? Or will there by different, better ways to help children learn what they need to succeed in life?
I don’t know the answers. My goal is to identify the early uses of new technologies and bring them to light so we can all figure out which educational primitive will work. For example, my latest podcast episode highlighted the role of voice AI in helping grow early literacy. Students in many classrooms - but not all - are able to speak and have a computer listen, and in some cases this leads to large learning gains. Does this mean that our future classrooms will further incorporate AI voice agents that can listen and respond to students?
Perhaps - we will see and learn as these gain adoption over time.
What we’ll cover
This newsletter will cover progress in a few key categories:
Research. We’ll cover recent educational research into efficacy of AI and look at what types of novel use cases are being studied.
Schools. We’ll talk to teachers and students in schools around the United States and internationally to learn how they are using AI tools. (If you would like to speak with me, please sign up for an interview here)
Products. I’ll analyze AI education products - from established players as well as brand new startups.
Themes. Finally, I’ll tie them together into the essential themes - the Base Classes that are driving student learning today and in the future.
Throughout this all, I will link to other thought leaders that I respect and pull from. I would love to have you along on this journey. Please subscribe, share, and send your feedback.