EdTech and the Followup Question

Every EdTech pitch deck since 2010 has promised personalized learning at scale. Most of what actually shipped was video lectures with a progress bar bolted on top — content delivery dressed up as pedagogy.

The genuinely new part isn't another platform. It's that whatever now sits next to a student can hold a conversation with them.

The followup question

Ask a kid why they got a problem wrong, and a good tutor doesn't just mark it incorrect and move to the next one. They ask a followup: what were you thinking when you wrote that step? That one move — the followup question — is most of what separates a good teacher from a mediocre one. It's also exactly the thing software has never been able to do, because it requires actually understanding what the student meant, not just pattern-matching their answer against a key.

Language models can do a rough version of this now. Badly at first. Then less badly. That's the real shift, and it's worth taking seriously precisely because it's unglamorous compared to "AI tutors for everyone" headlines.

The risk nobody's pricing in yet

An AI tutor has infinite patience. It never gets tired of the same question asked five different ways, never sighs, never has twenty-nine other kids waiting for attention. That's a genuine advantage — and also a setup for disappointment.

Kids get used to that kind of patience from a machine, then walk into a real classroom with thirty students and one teacher, and the classroom loses the comparison every time. That's not really a technology problem. It's a staffing and funding problem wearing a technology costume. No amount of model improvement fixes a teacher-to-student ratio.

This is the trap worth watching for as more schools and parents adopt these tools: mistaking "the AI is patient" for "the underlying resourcing problem is solved."

What EdTech actually needs to be honest about

EdTech doesn't need another app with a mascot and a streak counter. It needs products — and vendors — willing to say plainly which problem the AI actually solves (patient, infinitely-repeatable, one-on-one Socratic followup) and which problem it just makes easier to ignore (a system that was underfunded and understaffed long before any of this existed).

The tools that will matter aren't the ones with the flashiest demo. They're the ones built by people honest about that distinction from the start.