How AI is going to disrupt education

I almost titled this post: “Why Cicero would be a master technical engineer in the AI era”, but I was afraid people wouldn’t take it seriously. I still think it’s the best way to summarize it.

Given that our startup services the higher education market, I have the privilege of working with educators almost on a daily basis. AI introduced questions and some “discomfort” very early on (around 2022 when ChatGPT was launched), but I feel like this year the tension has reached a new maximum, and I’m starting to believe that education will be completely disrupted in the next 10 years.

And not in the way most people think. The conversation tends to fixate on students cheating on assignments, which I find to be the least interesting part of all of this. The two main points of disruption that I see are:

1. The technical topics we teach are shifting: some are becoming obsolete while entirely new ones are emerging that we can’t yet name.
2. The very goal of universities needs to be redefined: we may need to abandon the industrial-era model of knowledge certification and return to something more profound.

1. The topics are shifting (and we don’t know where to yet)

Every time humanity introduces a new abstraction layer, certain skills become less relevant while new ones emerge. Think about it: in the 1970s, a serious programmer needed to understand assembly language, pointers, and manual memory management. By the 1990s, languages like Java had abstracted memory management away through garbage collection, and entire generations of developers since have built successful careers without ever touching a pointer. Today, most software engineers work in Python or JavaScript and would struggle to write a line of assembly (me included). The skill didn’t disappear from the world, it just stopped being essential for most people. The same pattern played out elsewhere: logarithm tables became irrelevant once calculators arrived, and mechanics traded their deep knowledge of carburetors for diagnostic software once cars went digital.

AI is creating a new abstraction layer and most of the topics we’re teaching today will become obsolete. Is it still relevant today to ask a student to memorize the syntax of SQL? Do they need to know what a Primary Key or a Foreign Key is? We don’t know, what we do know is that the skills required are “shifting” towards higher levels of abstraction.

But here’s the catch.

Every previous abstraction created new problems we hadn’t anticipated. High level languages allowed us to stop worrying about low level memory management and the assembly language, but that meant that coding became more “accessible”, which meant more people were coding and the codebases became larger. With more people writing software, and the codebases getting bigger, we were introduced to completely new problems: agile methodologies to manage software teams, software architecture, new programming paradigms, etc.

So the question is twofold: what topics/skills are important to teach students and which ones will become obsolete, and how can we anticipate the new problems that the introduction of AI will create.

Moreover, is AI just another abstraction layer like the ones we’ve navigated before, or is it categorically different because it is abstracting across all human domains simultaneously: writing, coding, reasoning, analysis? For the first time, we may be abstracting away thought itself.

Which takes me to my second point.

2. The goal of the university itself is changing

This is the part that keeps me up at night. It’s not just what universities teach that needs to change, it’s why they exist in the first place.

The modern university, shaped largely by the industrial revolution, is built around certifying knowledge: you take exams, you prove you know things, you get a degree. But this model breaks down completely when an AI agent can produce correct outputs without genuine understanding. Give a student with zero SQL knowledge enough time and tokens, and they’ll likely solve the problem. So what exactly are we measuring anymore? What are we evaluating?

The shift may require returning to something older than the industrial model, closer to how ancient universities operated. Back then, education centered on rhetoric, debate, argumentation, and reasoned judgment in real time. Ironically, those ancient skills (asking the right questions, evaluating arguments, detecting flawed reasoning) are exactly what we need when working alongside AI. We may need to teach less about what to know and far more about how to think: how to validate an AI’s output, how to spot hallucinations, how to reason about whether an answer actually solves the problem at hand. In a strange twist, the most futuristic education might look surprisingly medieval.

The human problem

Aside from the philosophical and pedagogical questions raised above, we’re also facing a very “human” problem: the tension between the overwhelming popularity of AI and the reluctance of educators. Whether it stems from fear, pride, or simple inertia, I see professors refusing to embrace AI and refusing to accept that some of the topics they’ve taught for decades are quietly becoming irrelevant.

This is not the first time we’ve seen this. When handheld calculators became affordable in the 1970s, math educators met them with fierce resistance. The arguments back then sound almost identical to the ones we hear about AI today: that calculators would make students overly dependent on technology, that they would hinder the development of fundamental arithmetic skills, and that they would inevitably be used to cheat on exams. But the most extreme (and frankly absurd) argument was that calculators would make students dumber: that pressing buttons instead of working things out by hand would atrophy their brains and leave them incapable of thinking critically. That part, at least, was clearly wrong. Yes, most adults can’t do long division on paper anymore, but the world didn’t end and we didn’t become dumber.

What had to change was math education itself. We stopped measuring competence by how fast you could multiply three-digit numbers by hand, and started measuring it by whether you could set up the right problem and interpret the result.

The educators who resisted the calculator the hardest weren’t villains, they were defending something they had spent their lives mastering. Watching your expertise become optional is genuinely painful, and I have a lot of empathy for that. But it doesn’t change what’s coming. The professors who will thrive in the next decade are the ones willing to ask themselves an uncomfortable question: if my students can get the right answer without me, what am I actually here to teach?

The open questions

This post serves as a spark to get me thinking, but the questions remain unanswered:

• What topics will be relevant for students to “learn” in the AI era? Which ones are going to become obsolete?
• Will this be just a new “change of abstraction” as we’ve seen multiple times before? or AI imposes a more profound change of paradigm?
• How can we evaluate students? Is it enough to give them a problem and an objective, or we also have to evaluate their use of AI (the tool)?
• What will be the role of the university? How is it going to adapt to a less “knowledge-oriented” paradigm?
• How will people take this and adapt to this? we’ll be able to have smooth transitions or it’ll be 10 years of pushback and tensions?

I don’t know, I don’t think anybody knows, except maybe Cicero?