As a teacher, this is obvious. But when you ask a student 'do you want to learn?' they don't actually understand the question.

disagree- i used it for the last few years of coursework preparing for exams. If you give it context (i'm studying for an exam, do not give hints unless asked, your objective is to help my understanding) it is incredibly helpful. This isn't advice, it's a company going out of business.

They rule for maths. I had an extremely good experience going from high school trig through multivariable calc with them over a year. Best math teacher I ever had.

I don't know if I agree, if I ask a physics question it will show me the process to get to the answer, and it's really nice to have something to talk to with infinite patience to dig down into steps you don't understand. I learned a ton about implementing machine learning from using AI models. Obviously that takes a level of curiosity and effort on the learners part, but with a pretty minimal system prompt I think you'd get a pretty great tutoring system.

My daughter worked on learning for various classes by building her own plugin and skills for tutoring that uses various techniques of teaching without revealing answers, giving advice on homework questions without disclosure, etc. She overlaid with the personality of her stern gymnastics coach. It was surprisingly effective and I was pretty impressed at 11 she had the sense to not go straight to the answers. I think the “oh the kids” hand wringing is always overblown with kids - it’s the adults that don’t adapt well.

Students have always had the curriculum and standard study material, it was always possible to study alone, but the teacher mostly provides an interactive gap-bridging layer, and the classroom setting provides some level of obligation, I could be wrong but I don't know why AI can't fulfill the interaction layer part.

The ChatGPT web app, though apparently not the desktop app, does have a study mode where it will not give you the answer but will help you reason through the problem like the world's most knowledgeable, most attentive teacher.

I’ve found LLMs to be very good at teaching. Better than I expected.

You have to tell it what you want out of the session. A kid could just type in the question and get the answer, but that’s not tutoring.

When I use them for learning I tell it my goal, summarize my current status, and then tell it to teach me incrementally with a quiz after each step. You can get fancy and have it reach back into history to throw in extra refresher questions.

It uses very few tokens compared to a coding session so you can use the highest model for this and the usage is negligible. Using the top model available to you is important to keep things accurate. I wouldn’t have trusted this back in the GPT 4 or Sonnet days, but with Opus, Astra, or Fable it’s great.

For longer sessions set up a directory and have it keep notes about your progress, problem points, and what was covered in the past.

A calculator is a tool that replaced the slide-rule or Abacus - no one contended the Abacus replaced teaching or teachers. Ditto an iPad which is effectively just a fancy calculator in this context.

Similarly Youtube didn't replace teachers - it afforded remote learning and correspondence courses a new medium for transmission, recording, and playback of teaching.

More to the point none of those were trained on a corpus containing the history of human pedagogy, were capable of inference, or of generating dynamic learning resources based on best-practice models with inherent self-correction and pedagogical reinforcement mechanisms.

This time around, AI being able to do student homework means that AI is replacing students not teachers.

You're basically describing the established pedagogical field of Problem Based Learning, which is around as long as personal computing.

https://en.wikipedia.org/wiki/Problem-based_learning

In PBL the teacher adopts a constructivist approach - supporting, guiding, and monitoring the learning process as opposed to delivering a defined curriculum via lectures or seminars.

Agentic AI is arguably a better fit for this as it can scale infinitely to monitor student progress, offer timely feedback on design choices, and adjust difficulty dynamically based on how the student interacts with the problem.

Your main contention is one I'd broadly agree with - that use of Agentic AI for tutoring abdicates intellectual agency, allowing the AI to perform the critical synthesis and analysis that the student should be doing themselves for learning to occur.

The solution, however, is clear: guardrail the AI to promote core pedagogical principles rather than replace learner cognition by mandating things like instructional friction and the strategic fading of AI support as students build competence.

As explored in the paper I link below, by balancing proactive scaffolding with necessary and beneficial cognitive challenges, AI can preserve the productive struggle essential to effective learning.

https://arxiv.org/html/2606.04543v1

(A final thought: have a look at the current field of companies who specialise in turning training materials into engaging interactive experiences for the likes of Compliance and Security training. This is a corporate field now being automated and democratised for use by the average student in Academic pursuits in an on-demand fashion. Appropriate AI tutoring is basically the same concept.)

Alpha shows the results they do because they charge exorbitant rates (up to $75,000 / year depending on region) to staff the school with multiple individualized tutors per student.

It is a for-profit product which is sold to parents and state-level customers more than it is a school.

Slightly creepy, seems harmless now but future without supervision you don't know what they're teaching or what is going on. I would agree it's a double egged sword.