AI has been a fundamentally disruptive force in education, exposing some of our current system's weaknesses, providing new tools and new approaches to learning, and promising yet further disruption in the year ahead.

So, to document this disruption, gain some insights into it, and get a heads-up on where things might be headed over the coming year, I've interviewed an expert: ChatGPT 4o (June 2025 version). Check out the full interview here.

And/or for those who like listening to podcasts, here's an AI-generated podcast analyzing and summarizing our interview!

Final note: As usual, AI's analysis of AI tends to be relatively optimistic about AI, for some strange reason. I'm actually a little more cautious and skeptical, which comes across a bit better, I think, in the text of the original "interview": my goal is to promote understanding of AI - understanding of both its strengths and its weaknesses - and to get us all thinking about how to best use it, or even if we want to use it at all!

One of the first uses for the new Large Language Model (LLM) AIs, after the generation of human-readable text, has been the generation of computer code. This makes sense. There is probably as large a body of "open source" code out there and available as text to train on as there is human-language documents, and programming languages are a lot more structured and a lot less ambiguous than human languages, making programming a task that LLM AIs are potentially very well suited for.

That being said, while there are plenty of pretty standard algorithms (logically structured programmable work-flows) that AIs can very easily reproduce, the logical, organized, and creative thought required to plan and create a large application has not been something any of the LLM AIs were capable of until fairly recently. And, while they're still not great at doing exactly what we ask of them (I still haven't managed to get an AI to translate my favourite TRS-80 Color Computer BASIC game for me!), a growing number of programmers are starting to embrace the chaos inherent in getting the AI to create complete programs, and they've even come up with a name for it: "vibe coding"!

The idea behind "vibe coding" seems to be, on the one hand, to craft a prompt good enough to get the AI to create an actual working program for you that more or less does what you want it to, and, on the other hand, to embrace this necessary "more or less" aspect of AI coding and, instead of trying to force the AI to do exactly what you want it to, to take what the AI has produced and tweak and/or just run with it.

Continue Reading