Somebody here posted about a Socratic method skill for Claude (or whatever your LLM of choice is). I’ve found this to be fantastic. Dropped in a dense PDF about Linear Algebra, Postgres (those are the two I’ve gotten into so far). It’s basically just a skill.md that reads, “Here’s a document, give me an opening statement about the material and ask me a question. As I provide answers, guide me to simpler or more complex areas of the topic. Something something Socratic method.” Works great.
Yeah. That was it. I remembered it wrong (you use chatGPT which I admit, has better voice interaction). I use claude code and type it all. Most of the time for walks/runs I just let my thoughts go whenever. But that's just me. I do really appreciate your post though. I wouldn't have the claude skill without it.
Spaced Repetition flashcards or flashquestions to progress through a game might be good; configure Anki decks to surface cards from when for example the user jumps to crush a bonus block.
Kaggle Learn has courses in ML, Deep Learning, ML Explainability, Python, Pandas,: https://Kaggle.com/learn
This is the Dockerfile that they install Python packages with on Kaggle Kernels, which is the Jupyter notebook product that Kaggle Learn is built on:
https://GitHub.com/Kaggle/docker-python
I've made an entire platform based on Socratic AI Agents that co-work with a person as method acting assistants in an office suite. https://midombot.com/b1/home
I'm not super familiar with this approach, but couldn't you just respond with an elaboration on how you don't understand a concept when you encounter it and guide the model in the correct direction?
That requires you, the student, to easily identify why you're not understanding and what concept among multiple ones is the primary or secondary blocker.
Haven’t found a model yet that really orchestrates in a way that reflects how learning works. Or at least, how I learn.
I can get them into different teaching/learning “modes” and configure those modes appropriately, but they don’t know how to switch between them in context and automatically, and they ALWAYS go tangential and ADHD mode after depressingly few exchanges before they need to reset. I’m talking like, 5? Just not good enough to really satisfy.
Huge improvements over time of course, but they seem to have plateaud for this particular issue.
I started learn linear algebra and some mid-level math stuff couple of years back. While I really wanted to learn I still struggled to find a good learning path and use a product similar to Duolingo for math.
Eventually I learned some, then stopped using it and then forgotten all of it.
But my point is despite if me really want to learn I found it quite challenging to find learning path that I trusted to be good start for me.
I also need to make the point; I didn’t have 5 hours a day, so this 30-90 minutes in a good day endeavor.
There are still tons of things to solve in this area, how to cross to chasm, a learning path that’s right for you and how can you be sure of that. Learning something in one good 60m instead of 5 different videos and 5 hours. Etc. I think these problems are far from solved.
I also want to get fit but that doesn’t mean I’ll show up in my gym to workout everyday. So even when you really really want to learn you still need to stay motivated.
The issue with sources like these is that they are inherently passive. They don't include exercise sets. While they can be good supplementary material, learning happens by doing, not by passively watching.
What you're looking for is not something to help your with your goals, it's something that sells you the idea that you can achieve your goals while seemingly be unwilling to put the necessary effort in.
What if it's simply impossible to learn linear algebra and mid-level math stuff by spending only 30 minutes a day? What if you can't get fit if you don't go to the gym?
It sounds like what they are looking for is a pathway. Khan used to have a really good one with the world of math pathway where you could go for 2 + 2 to Calculus in a set sequence of steps each building on the next. I've tried Khan a few times since they got rid of it and it just feels like you're all over the place and regularly missing skills it expects you to have.
They aren't trying to get fit by not going to the gym they are looking for a progressive overload program for math.
People want to learn at the last moment, when it is needed to know, and then they want to discard that knowledge until it is needed again. We need to implement both in our modern software: it needs to solve complex problems, and it needs to guide a human (or AI) operator in the subtle nuances of that complexity so their solution of the moment is good to last for the time being.
AI isn't that general purpose. If There is a chef course curated / tested with Jamie Oliver I'll buy that over a Claude chat. And by "I" I mean regular folk who are most of the market every SaaS and AI wants to capture.
This is exactly right. I started teaching undergrad CS about 18 months ago after 20 years wandering through other careers, and I quickly realized that students do not need better teaching. They need frequent, direct feedback and an environment that more or less requires them to commit to studying until mastery or to quit early and do something else.
Effective learning techniques have been known for decades, but universities have -zero- incentive to improve teaching quality. There is no standardized assessment infrastructure to compare graduates across schools, or to evaluate the effectiveness of teaching interventions at scale. So any effort put into teaching quality is invisible and thus worthless to a university.
Every attempt to make high-quality instruction available for free has been largely irrelevant to the unwashed masses, because universities have a monopoly on credentialing. The self-starters can learn anything for free, but they were never the problem. It’s the middle 50% that we have utterly failed.
I believe a huge part of the failure is the lack of teaching anywhere, the general idea of how to manage discussions of disagreement. Because that is not discussed or taught, any topic with disagreements triggers emotional reactions in just enough people that no discussions of any progress can be made in such areas.
When the "middle 50%", as you call them, find they need help or really anything beyond what is immediately available they find they cannot articulate their need, nor can anyone around them understand them in their requests. They are propped up with an incomplete script and just left to flounder around, and because they cannot articulate their situation larger society is allowed to ignore them. Which is a completely manufactured situation.
I think learning from another human being live who can understand your confusingly articulated questions and who can adjust their answers for you, another human being, that's great.
I agree, at least in that we need more human contact, not less.
But I will say Gemini / Claude have both been surprisingly good at understanding my half-assed attempts to put a complicated question into a coherent sentence.
What obstacles do you tend to run into? I'd be inclined to use a Chinese model for that, since Claude was presumably RLHF'ed primarily by native Anglophones.
Have you tried a Chinese-trained LLM for learning Mandarin, and if so, did it help?
>Too costly. Now its viable with AI to do it cheaply.
Where is the data that shows this sort of technology beats normal pedagogy?
From a study I read that came out last year, AI tutors are at best no better than a teacher, and whats worse is that AI tutor use caused many students to become dependent on them. These students could not recall information or perform nearly as well without them.
This is just another example of a long line of behavior from software engineers thinking they are god's gifted people, who can waltz into any industry and revolutionize it overnight with an application.
Scorm is a zip file with html files to structure content any way you want with a minimal amount of mandatory standards.
It was never meant for assessment, nor is it secure. The creator of scorm said to not use it learning had to be secured.
It’s not a system or a framework in any meaningful or standardized sense, and more of a format for compiling content that can be copied between different places.
I just took the conceptualization of it away from the sufficient experience I have with LMS to do the same with it:
Learning management systems are rarely about learning. Too often, they're first about about managing learning for the learning manager, and teh priorities of the learning manager (maintaining compliance) and that's what the software and sales process aims for.
The learner, actually learning, or being a central focus is a second or third class citizen in such tools.
Yep. Literally every EdTech idea is a rehashing of the same flawed reasoning: we just need better monitoring and metrics to manage everyone’s learning process! No. Learning comes from repetitive deliberate practice in the zone of proximal development. For essentially all knowledge-based skills, what’s missing is the core foundational skill progression that leads to actual mastery.
How many different personal training styles do you see for competitive sports? There’s variation, but only at the margins. Nobody deviates from the essential training core because doing so leads to poor performance and competitive failure. And yet we have a total free-for-all in knowledge disciplines where we just throw untrained PhDs in front of a classroom and assume they’ll produce results. Oh, and we also provide zero tangible incentives for them to improve learning outcomes.
How to teach, and how to learn, and how to learn how to learn are for sure different things.
With post secondaries, one thing I see is the obsession with pedagogy, which makes sense on one hand, where teachers who teach K-12, learn how to teach children, for the unfamiliar - peda is children, and gogy roughly means "how to lead".
If our teaching is rooted actually or inaccurately using a word that is rooted in a practice of how to lead children, it doesn't surprise me that the learning experiences of non-adults, including young adults, are so hard.
I've got to work in K-12, Academia and Industry training/certification, and where they work, they do not share as much as each of the categories would like to think.
Thus, when someone takes a K-12 framework or approach and applies it in industry, it's feasible that it might not work, but the takes on why seem to get ignored.
Fundamentally skills and the ability to use your skills change and evolve and fixes curriculum that takes 1-3 years to change one sentence doesn't really work in more and more places. This is more a fundamental change than a short lived irritant.
You don't have to see the value - the learners have to benefit, being the majority, they usually won't be setting it all up like you are implying on their computers.
Is it? People want to learn things they think they need. Is there a growing disconnection between what they think they need and what they actually need?
I mean "People who want to learn will gravitate to learning systems." means that this is the obvious target market for a learning system. Since there are lots of those out there you win on execution.
>If you can find a way to pull in people who don't want to learn, that has value.
Most companies try to address a known market and not create one.
At any rate you have two types of people who don't want to learn:
1. people who don't want to learn and have to anyway, for some reason (generally work related)
2. people who don't want to learn and have nothing forcing them.
Number 1 is a potential market, number 2 seems hopeless.
Most people don’t want to learn, they want a credential. The real hard problem here, that has been sidestepped for decades and decades, is how to gate credentials behind standardized skill assessments rather than branded gatekeepers. Stanford, Harvard and MIT can do whatever they hell they want in the classroom as long as they continue to skim the cream of each generation and no objective assessment exists to demonstrate that they contributed nothing above what their exceptional students probably would’ve done in four years anyway.
As a daily user of Anki and Math Academy, I'm very excited about how software can help us learn.
Now, with AI, we can really build tools that were not possible before. I built for myself AlgoTutor[1], a CLI tool that teaches Go developers algorithmic skills and concurrency.
The recent Understanding floating point[2] from Oz Nova is also a type of tutorial I want to see more often.
Math Academy is doing it right. Their neat synthesis of spaced repetition and the knowledge DAG is slick, but honestly the extremely fine-grained skill decomposition is what’s missing in virtually all of higher education. I’m trying to follow MA’s example as I put together curricula for my undergraduate CS courses.
I'd never heard of Math Academy. It seems interesting, but quite expensive ($49/month for software which is in beta). Would you say it's significantly superior to something like Khan Academy (which is free) to justify paying that much?
It is indeed expensive, especially considering my current location(Romania).
I have a rule of doing 20 XP per day, and I kept that rule for more than half a year. Maybe the price is also a factor, but I never managed to keep that pace for Khan Academy when I tried it. You can check their current courses here: https://www.mathacademy.com/courses
I think the topics and curriculum are more serious, and I would definitely recommend it if you have kids or are in highschool/uni. I'm 37 now and only do it to keep my brain sharp and build my mathematical intuition and I think is great for that due to its focus on practice and spaced reviews.
I think they have a 1-month trial, and you get your money back if you cancel within 30 days if you want to try it.
I tried Math academy twice and I cannot understand the hype but a lot of people seem to love it so maybe it's just not my style of learning. Each topic seemed to contain too little information to actually grasp it. It was like
Trig 1 -> This is a triangle and trig is about triangles.
It is vastly superior to Khan. Had my kids go through both, so can compare.
That said, on higher levels there's some cruft that trains mechanical skills, and that's on subjects that are needed only in the scope of the US system. For Europeans these should be optional, alas they are not and this adds a lot of tedium and hassle to an otherwise rather frictionless process.
I do think Andrew is the right person doing this. Someone mentioned that the return of financial outcome is not good in edTech, and this is exactly why backing by Coursera is a huge advantage. Education-related product is slow, and it needs patience.
The concern, though, is that how to achieve those 3 thesis. Do they train/fine-turn their own model? according to my own experience, the current models are not well trained in personal education context, and it is either too textbook-like or too task-driven.
Yes, Andrew Ng is a great educator, well positioned with Coursera to deliver this, and personalized education (with exercises and feedback) is an excellent potential use of AI.
I guess the real challenge would be initial and continual assessment of each student in order to select the next instruction that they are ready for, but with "LLM as judge" that doesn't seem such a problem, and student feedback can be used to continuously improve these recommendations/selections.
Presumably you would still want to have a human designed curriculum, with the AI being used as the teacher - assessing how each individual student is doing and what they need to be taught next.
I remember reading that Education is a trillion dollar market, so i'm not sure about the financial outcome being poor, only the state of how that money is spent and obligated can differ greatly.
The biggest success that people usually point to in EdTech is Duolingo. They were founded in 2011 and IPO’d for about $6.5b in 2021. A big success by any measure. But that’s the pinnacle. Compare that to DoorDash which is an $80b company, let alone the trillion $ AI labs and you’ll see why the largest VCs hardly touch the space.
- K-12, Academia/Colleges/Universities, Industry/Training, Adult learning, increasingly experts teaching directly, and more.
Most people aren't aware of the differentiation of the categories because they get to work inside one for the most part.
The learners in each category are different.
Duolingo had it's share of naysayers and doubters, but it did provide it's own take on a learning outcome.
Your post does make me wonder about the existing spend in Education, and how much could/should be EdTech with today's capabilities, or even the capabilities of 10 years ago, and what looking at that with a lens like yours might show, if anything.
There’s a reason services like GitHub Classroom Replit’s education offerings shut down. Serving educators can be very hands-on. Education can have lots of different requirements (every teacher needs this one little feature), often legal issues (student privacy), and dealing with universities and governments is awful and bureaucratic. Most companies will find it easier to make money elsewhere.
The dirty little secret of EdTech is that there is no money in improving learning because there is no way to capture the value. For all but a handful of highly specialized and insular professions, there is no scalable infrastructure for objectively assessing skill mastery. Most learners don’t want to learn for its own sake, they want a credential that proves their value to potential employers. All of the value in learning accrues to the credential, and everyone in the educational establishment is quite happy reserving that privilege for the accredited universities because they can just bundle more and more products and services into the tuition sticker price without ever having to justify it. Oh, and the student loan racket ensures that enough students are price-insensitive that they’ll never look for alternatives.
They're just outside of K-12 and Academia, who don't necessarily keep the gap close to the real world and it's needs post-graduation
The value in learning is what you do with it, the credential can be a part of the journey but not always. People were definitely about the brand, and it's value, of the stamp of a piece of paper, however given the experience of recent graduates the past 5 years, it's revealing itself to not be an aberration that's going to give way to the 500-1000 years preceeding it.
Yeah, that was my first reaction to that headline too. Given a second thought, he probably means the opposite of "replace", so in this context, "maintain people." maybe? I know marketing headlines shouldn't be interpreted literally, but it tripped me up at first.
I had the same thought. Then I thought, maybe "replace" is what will become opposite. Antonyms of "replace" are
hold
keep
precede
break
damage
deprive
forfeit
lose
penalize
remain
take
worsen
This is one of weaker Asimov's works, in my opinion, as it merely projects pre-mass education era teaching into the future. Conflating mass education and socialization is pretty myopic when writing about the future with unlimited intelligent robots. You don't have to be taught in one-size-fits-all way to have shared experiences, it just happens to be done that way due to the scale, out of necessity, and many education systems and experiments since the 90's are trying to combine individualized learning with socialization. Unfortunately many post-WWII sci-fi predictions turned out to be self-fulfilling, because a lot of people got conditioned on them and can't imagine breaking the lore which is now a part of their lives.
Care to enlighten the rest of us how this is at all related to the submission? Is it because it's AI it is "destroying human everyday life experience"? Seems this company is trying to find a healthier (and better) way of using AI, unless I misunderstand something?
If said experience is subpar, e.g. when it happens with 30+ student classes when 80% dgaf, there's no harm in offering an alternative to those who wants it.
And the rest can continue enjoying everyday human life experience.
It is. I just wish people would throw a plain text page with a paragraph instead of this because you bet I am just scrolling through it without reading as soon as Claudism appears
The home page is entirely written by AI. Riddled with em hyphens. More should be placed on the home page at least..
But really what a wonderful mission. Imagine if they succeed and end up teaching millions and allowing people elevate their home towns/countries. Crazy impact.
This looks like another investor play to save Coursera and somehow have a vector (pun intended) to help raise its valuation.
Coursera and Udemy already offer AI learning tools, so why create a separate company and commit $100 million? It looks like the product, commercial relationship, ownership structure, IP rights are still unclear.
I’ll give him the benefit of the doubt and hope for the best.
My wife has been a high school math teacher for 25 years and has seen waves of tech luminaries give their visions of how tech will radically improve education and learning. She can’t point to any startup that has had a major impact in improving outcomes. I wish I could point her to something that really will have an impact in, or even out, of the classroom for her students.
That's not unique to tech startups. Almost nothing in terms of explicit methodology has any effect on education outcomes. The literature on education research is of abysmal quality too.
I got to see Luis von Ahn the inventor of CAPTCHA discuss his latest crowd sourced venture, Duolingo, which I'm sure everyone has an opinion about, back in 2013 or 2014 at The LAB Miami in Wynwood.
At that point they had ~24 employees. He said his interest was crowdsourcing and thought most of the engineering effort would go toward the crowdsourced translation, but at that time, it was being handled by just two employees.
They brought on two prominent consultant researchers, leaders in the field of language learning and linguistics. The first question they asked the consultants was which part of speech should they present to the learner first. The consultants didn't have an answer. They decided to dedicate engineers to collecting data in order to answer that question.
Very cool to hear that you were at the LAB Miami in 2014, Around 2016, I was trying to organize a group of folks to build a variable spaced repitition application that tied to a off the shelf eeg, to figure out the optimal time to learn at one of the local hackathons.. We had a whole team of folks, some of the Flomio hardware people, and other local startup folks. Never was able to build the prototype at a hackathon, but always thought there is a physiological element to learning, and ed tech in general.
We know that mild stress can narrow focus and help learning. "Neural connections do form when you struggle." [0] A little stress is not only good but necessary for learning while a lot of stress causes a breakdown. The trick is being able to judge the stress level of the student and every person is different.
The idea of using some type of bio metric data from EEG or blood pressure sensors will help determine a little struggle from a lot of struggle in formative assessment.
My understanding (in hindsight) is that success of Duolingo had nothing to do with “crowdsourcing” - is that correct? (Beyond maybe stories for fundraising).
It doesn’t seem to have a lot to say about “ed tech” - popularity of apps seems disconnected to popularity
I don’t understand what you mean by “we don’t know”? Are you saying tech entrepreneurs don’t know education? I agree with that. Or are you saying we don’t know what problem to solve?
My interpretation of “we don’t know” is that the space of possible instruction methods is huge, and we don’t know what works best. It is easy to think of it as a kind of optimization problem, but that line of thinking quickly runs into a bunch of practical obstacles: different people learn differently, studies with good experimental design take time and resources so it is slow to iterate, and good teaching seems to require deep insights into student motivations and incentives, aspects of teaching that are hard to quantify. We also don’t have great ways to measure outcomes; grades are weakly predictive of future success.
The best teachers seem to combine deep thinking about how to teach with intuition and charisma. The ingredients are hard to replicate.
I do think that this is a nut that we will eventually crack, but it is proving to be a very tough nut.
There is not a huge market of people who actually want to learn something on their own motivation. Get some piece of paper, yes. So the huge market will only materialize if the degree granting institutions adopt something like this.
May I ask why? If by "not the same" you meant: AI is not at the level of a good human educator, it is a fair argument - but not many humans can afford to have a one-to-one human educator either.
Perhaps I am a bit more optimistic, but if AI could give me, let's say, 90% of an average lecturer's capability, at 10% of the price or even less, that is probably extremely valuable. (For one, I would certainly be interested in learning new things that otherwise wouldn't make sense financially)
Moreover, there are areas of expertise that are very hard to reach without special help. I have a PhD in computer science. Earning that degree was a major time commitment, and it gave me close contact with one of the few experts in a niche area. I absolutely loved the experience.
But I did not learn everything I wanted to know. And I can’t afford to take a time-out to go to grad school again for a second PhD. On the other hand, an LLM, in some sense, has access to the world’s knowledge. With some patience and persistence, you can mine almost-expert-level instruction from one.
When ChatGPT 3.5 came out, I excitedly tried this for a topic that I WAS an expert in. The results were discouraging. The model’s answers seemed to reflect the general misunderstandings that people had about the topic. More recently though, I had to review a paper in the same area, and I used a late-model ChatGPT to help check my work. It was an eye-opening experience because it was no longer confused. And it found longstanding misunderstandings I had, buttressing its answers against my skepticism with citations to original work. I came away very impressed. This kind of AI “rubber duck programming” is my preferred style of use now. I used it just today to help me learn an area of statistics I have always been fuzzy about. This approach definitely requires some careful prompting, but I am optimistic that AI tutors will one day be a real thing. My only worry is that people lose the ability to understand what makes an answer a good one and why we should care about good answers.
>And it found longstanding misunderstandings I had, buttressing its answers against my skepticism with citations to original work.
Did you check those citations to make sure they actually contained the information they were being cited for? Because I've asked difficult questions too and have had mixed results. I would never trust an LLM as a teacher. They make stuff up all the time, and when you're learning you're in the very worst position to be checking the information you're given.
A lot of knowledge in such niches has never been written down. Papers don't contain it. Tacit knowledge, it's in unpublished config files and unpublished experiments that some prof or PhD student remembers, all the things that are not "publishing-worthy" (because not flashy, not novel-sounding enough, or a negative result etc) but crucial for the method to actually work. Most of expertise is like this and this is why spending years in a research group, talking to the researchers, participating in their culture, understanding their thinking over years (or even just months) can be invaluable. There are also many things that people only say after a beer, after you know each other well. Claude and ChatGPT haven't ingested any of this. They ingested the sanitized surface version, the press release / advertisement version. Because papers are basically career marketing and advertisement mostly. The real expertise is guarded and hidden.
This is of course about the PhD level and beyond. For regular university-level education, where the goal is to pass the usual exams, ChatGPT and Claude are more than enough. But for real-world "doing", not (yet).
I just fundamentally believe that AI cannot replace the human aspect of connection. A teacher does more than just teach, they also connect with the students and form relationships. You could argue we may end up with some form of “AI” connection though if that’s a good thing and what is its long term effects on us as a society we are yet to see. Regardless, what I mean to say is I don’t think AI to Human can replace Human to Human as a whole, that’s my point. The marketing copy feels deceptive and not something that speaks to me, I wouldn’t consider that to be same. This feels like it’s marketed towards MOOCs but I am not a fan of education as a product like that either, I come from a place where education is free and this idea of gatekeeping knowledge and forming products around it just doesn’t sit well with me. This just feels like a way to make more money and marketed towards that narrative. You’re right it lowers the cost, but at what cost? MOOCs already didn’t accomplish much the way I see it, I saw much better results from self made paths with things like YouTube or free lectures and/or self made teachers who go above and beyond. That said, I think there is a role for AI in education and I do think it belongs in there, alongside a human educator, in some form.
I didn't have "replace" in my mind. I 100% agree with you that AI cannot replace human educators, but I am hopeful that as a supplementary tool, AI can achieve a lot.
On your point of MOOCs and YouTube lectures: I agree that the latter is sufficient for me, but there's nothing stopping AI from creating something similar, or even better: lectures tailored to my specific interests and needs. MOOC failed for me exactly because it wasn't "for me" - I might as well try and find a YouTube playlist.
These are all tangential, of course. I don't know how this specific thing would turn out.
I believe if you need a human tutor to understand category theory or Lean then you don't have the requiste math/CS background, and a human/LLM tutor won't be able to meaningfully teach you anything.
Note I said "meaningfully" - something with LLMs that deeply concerns me is that they provide edutainment, and people actually think they're learning something. One of the comments on this thread mentioned giving a "dense linear algebra pdf" and having an LLM use the Socratic method. I guarantee they would have learned more if they read the PDF. I strongly doubt they learned anything at all with a bunch of silly "Socratic" questions about linear algebra. But it felt like they learned something!
" if you need a human tutor to understand category theory or Lean then you don't have the requiste math/CS background and a human/LLM tutor won't be able to meaningfully teach you anything."
No, it's an honest thing to say. If you're not able to read a graduate-level textbook and teach yourself category theory then you don't have the mathematical maturity to understand category theory. You need to start smaller: ideally abstract algebra and point-set topology, or set theory and mathematical logic if you just want the CS applications. Likewise with Lean. Of course there is a place for a good human instructor. But mathematical maturity must be developed the hard way.
I'm working on a much smaller idea, but so far it's just me using it for grad school, and a friend using it for his master ciccerone prep. It has been really useful in the class I used it for while building it. Looking forward to playing with more as I start my fall semester(UIUC CS 425, Distributed Systems).
I need to get some model pricing evals in place to see if I can use open weight models to get it cheap enough, since it's just me and not VC money. The real product potential seems to be getting reusable materials like cert prep, continuing education for careers that require it, and the like. But then you're fighting incumbents that fiercely control their moats.
I'm curious: what looks awesome? The website was created with Claude. There's no detail about what the actual product/service is and how it will work. Everything is a big idea/promise with no specificity.
I've been working for a few years now as a private tutor teaching proof-based math to adults. Maybe this should give me an incentive to join in on some of the cynical replies this has gotten, but that's not actually how I find myself feeling. I love teaching, especially in the one-on-one format, but it really doesn't scale; not everyone can afford a human tutor and there probably aren't nearly enough of them to meet the possible demand.
If a product like this worked well --- and I agree Andrew Ng's involvement seems like a positive sign --- then that would be great! Something I would wonder about from my past experience using LLM's for technical subjects is how to replicate a particular side of the student-teacher relationship. Sometimes a new student of mine will need to be told that they don't actually understand something as well as they think they do, and that they'd be better served by going more slowly than they might have wanted to. This "talking you down gently" conversation is one I'm happy to have (and sometimes it's actually me who's wrong!) but I'd be interested to see whether and how well an LLM run by a company that really wants to retain customers could pull this off.
It's hard to say without an actual product, of course. I'm looking forward to seeing what they come out with.
To help anyone else who hadn't also previously heard of Andrew
Andrew Yan-Tak Ng is a ...computer scientist and technology entrepreneur focusing on machine learning and artificial intelligence (AI)... Ng was a cofounder and head of Google Brain and was the former Chief Scientist at Baidu.
Ng is an adjunct professor at Stanford University... Ng has also worked in online education, cofounding Coursera and DeepLearning.AI...
Hopefully the app itself won't look AI-generated like the design of the website. This one isn't so bad, but has hallmarks of Claude designs. Nowadays I keep getting shocked to find severely sloppy designs when browsing through Y Combinator startups. For me, this is a signal that the people behind the product don't care much.
So the founder of Coursera is starting a new venture backed by Coursera ? That sounds like the Musk playbook.
I don't really get how this is new or even "AI". Can't you already ask Claude to teach you something ?
"Patiently stays with you until you've mastered new skills", that's so nice. Maybe in the future it will be the other way around, your computer will complain about how slow the user is :)
I think this is a great space because almost any AI educational product can be measured and evaluated by the end user. For example, does the user improve from using your product and do they also believe they do, and do they also LIKE your product? If not, then they have every right to seek out the perfect AI educational product, of which, I expect a variety of.
Personally, I believe the last 40 years or so of academia has failed at creating the proper reactive curriculum for students, so I'm nearly positive anyone that enters the AI education space has a very good chance of proving to their end users that they can out-teach ... I don't know, the entire NY Board of Education, for example.
It is not bold at at all to say K-12 will be eradicated. Higher Ed not so much because many were always buying the status of higher education anyway, so the status in itself still has value as a product.
> It is not bold at at all to say K-12 will be eradicated. Higher Ed not so much because many were always buying the status of higher education anyway, so the status in itself still has value as a product.
It’s funny because I believe the opposite for both of these. Living with a teacher, I’ve learned that the most important objectives of k-12 education are teaching kids how to act like adults in society and behave around others. This is something I don’t think AI will ever be good at teaching.
A few other things, like learning to read and write, basic math, critical thinking are secondary, but important, but, I imagine not the best use case for AI either.
Beyond that however, you’re an adult equipped to take learning into your own hands with all the understanding and agency to make it happen. With the help of AI I don’t think there’s much you couldn’t learn, at least that doesn’t require human interaction (e.g. teaching).
Oh it will be even better than humans some day. Not any time soon, and definitely not when the sort of people training these things range from incredibly awkward around people to downright psychotic in public. But the infinite patience and persistence of coding agents in the form of a teacher for other things than code would be pretty amazing, but we're nowhere close to that.
Case in point, I would love to know who thought deploying a figuratively "demilitarized" sexbot as a teacher in NY town in 2026 was a good idea. And I say that as someone who loved the movie Weird Science. And I'm evil enough to admit that Van Halen's Hot For Teacher was the first thing that slipped into my mind reading the article below. But really? Who?
Ive actually started trying to build games, starting with Roblox, to eventually build a catalog for this same problem. 1:1 learning in the future is probably going to be 80 - 90% gamified because people like playing games. Spent the last 6 months traveling Asia and everybody was playing games on their phones...everywhere. The first game I released is Math Cadet, basically an adventure game to build up math skills for kids elementary and middle school. I have another one in the chamber thats more science related. Any feedback would be great: https://www.roblox.com/games/129132556457079/Math-Cadet-Adve...
I wonder what they will do different compared to Khanmigo, which wasn‘t particularily successful wirh full-on AI learning assistance. Support for traditional human learning systems?
"We're heads‑down building, and will have products to show by early 2027."
By then frontier models may be able to do this by prompting. Even now it is very easy to get one of the chat interfaces to keep quizzing you and adapt to your level, answer your questions, and in general behave how I imagine a tutor behaves. It lacks integrated stats/progress/gamification/long term memory but those can be added by a simple vibecoded app or are usually unnecessary for actual learning.
I hope they’re planning to build hardware so the learning experience can be close to pen and paper. Keyboards and screens are a distraction and the brain learns better with a pencil than with a keyboard.
I feel the exact opposite. I hand-write very slowly, so having to focus on that instead of the material is literally a distraction, compared to just typing out my thoughts.
The exception to this is if the focus of the learning is the writing in itself. Right now I'm learning kanji, so I need to draw them to focus on and memorize the shapes. That said, it doesn't have to be on paper. Dragging my finger on a tablet or moving a mouse are just as effective as using pen and paper.
>Handwriting activates a broader network of brain regions involved in motor, sensory, and cognitive processing. Typing engages fewer neural circuits, resulting in more passive cognitive engagement.
Yes, that's exactly what I mean. If I'm focusing on what my hands are doing I'm focusing less on the material proper than I'm trying to digest.
The idea is that focus and actually working all of your hand instead of just fingers helps learning the material more efficiently even if your own perception of it does not agree.
Maybe it works for some people, but it certainly doesn't for me. When I'm handwriting, especially in bulk, all I can think about it how much I hate it and how I can get it to end faster. I take fewer notes with more abbreviations that are later harder to decipher.
It's different if it's something like mathematical expressions, because those are never easier to type, but just regular sentences? Absolutely not.
Interesting, I can see where you're coming from. Key detail is the notes are actually not supposed to be read later, just written down - the act of writing by hand helps to commit their content to memory.
I get what you mean, and have noticed that effect on my own too, but the increased recall doesn't supplant the note, it just indexes it. I might hear something and go "oh, I remember making a note about that; it was something about magnets", for example, but I still need the note for the specifics. I haven't noticed any difference between typing and handwriting, in that regard.
There's a huge potential in personalized learning/education space. A big problem in education is that everyone learns different, and teachers / schools have to go by the speed of the lowest common denominator. AI can tailor content to the individual students most optimal way of learning. Whoever cracks that code will literally boost humankind
Not really, have you tried to do anything innovative working remote with a team? In any meeting half the people are checked out watching youtube. The collaborative whiteboarding experience is terrible. Working remote is good for plug and chug type work, but the team dynamic is an empty husk compared to working alongside others in the real world. It can be done if you absolutely have to, but don't try even pretend the effectiveness is anywhere close to being in person.
Yes, I've worked remotely for the last decade+ on many innovative teams and typically find them more engaged, autonomous, and innovative than most of the on-site-only teams I've also occasionally helped.
That would cover school breakfast and lunches for NYC Public Schools for only 3 weeks. Personally, I'd rather see investment in technologies that can give more than just one large school district a temporary improvement.
Sophisticated personalized education for the masses is finally within reach. Andrew Ng is one of the most significant leaders in ed-tech. I'm grateful that he's pursuing this venture and hopeful for the impact it could have on my family and many others.
Somebody here posted about a Socratic method skill for Claude (or whatever your LLM of choice is). I’ve found this to be fantastic. Dropped in a dense PDF about Linear Algebra, Postgres (those are the two I’ve gotten into so far). It’s basically just a skill.md that reads, “Here’s a document, give me an opening statement about the material and ask me a question. As I provide answers, guide me to simpler or more complex areas of the topic. Something something Socratic method.” Works great.
Probably referring to my post:
https://news.ycombinator.com/item?id=48938375 (100 up votes)
Direct link to the doc page:
https://pchalasani.github.io/claude-code-tools/plugins-detai...
This was in the context of how I use ChatGPT Live to brainstorm/discuss/learn while on a walk, but of course useful in other contexts too.
Yeah. That was it. I remembered it wrong (you use chatGPT which I admit, has better voice interaction). I use claude code and type it all. Most of the time for walks/runs I just let my thoughts go whenever. But that's just me. I do really appreciate your post though. I wouldn't have the claude skill without it.
https://github.com/pchalasani/claude-code-tools/blob/main/pl...
Khanmigo specifies the Socratic method as well.
These look like system and agent prompts and agent instructions: "Academic Assistant Pro.md", "Code Tutor" https://github.com/linexjlin/GPTs/blob/main/prompts/
Gemini Notebook does flashcards, study guides: https://edu.google.com/ai-gemini-notebook/
Spaced Repetition flashcards or flashquestions to progress through a game might be good; configure Anki decks to surface cards from when for example the user jumps to crush a bonus block.
"Ask HN: How can I learn to read mathematical notation?" (2018) https://news.ycombinator.com/item?id=18511718
"[Feature Request]: flashcards to advance in the game and learning mode" Issue #3054 ; SuperTux/supertux https://github.com/SuperTux/supertux/issues/3054
Resources for teaching ml and ai?
So I Ctrl-F'd for "Linear algebra" resources in my research log and found:
"Interactive Linear Algebra (2019)" (2021) re: Manim https://news.ycombinator.com/item?id=28173565
Latest Manim developments; interactive WebGL in browser: https://news.ycombinator.com/item?id=49093210
Math Academy has a Math for Machine Learning course: https://www.mathacademy.com/courses/mathematics-for-machine-...
Kaggle Learn has courses in ML, Deep Learning, ML Explainability, Python, Pandas,: https://Kaggle.com/learn
This is the Dockerfile that they install Python packages with on Kaggle Kernels, which is the Jupyter notebook product that Kaggle Learn is built on: https://GitHub.com/Kaggle/docker-python
https://github.com/bevibing/socrates-skill
I recently used this to prep for interviews - I got the job :)
I've made an entire platform based on Socratic AI Agents that co-work with a person as method acting assistants in an office suite. https://midombot.com/b1/home
I've never found this to work properly, because AI doesn't have the empathy to understand where you really struggling.
I'm not super familiar with this approach, but couldn't you just respond with an elaboration on how you don't understand a concept when you encounter it and guide the model in the correct direction?
That requires you, the student, to easily identify why you're not understanding and what concept among multiple ones is the primary or secondary blocker.
It's not easy.
Would you care to share them?
While the model itself is great at hiding the answer, the AI summary keeps giving it away, it's so annoying and AFAIK it cannot be disabled.
If you are using Claude code, disable recaps.
That's super interesting - will have to take a look into this!
> 01 Plans a path with you
> 02 Adapts to how you learn
> 03 Patiently stays with you until you've mastered new skills
1 and 2 any of the big AI companies will do that for you. I have Claude doing this for me for a few learning areas.
3 to me feels like the Duolingo bird.
IMHO, they are trying to solve the wrong thing. People who want to learn will gravitate to learning systems.
If you can find a way to pull in people who don't want to learn, that has value.
Haven’t found a model yet that really orchestrates in a way that reflects how learning works. Or at least, how I learn.
I can get them into different teaching/learning “modes” and configure those modes appropriately, but they don’t know how to switch between them in context and automatically, and they ALWAYS go tangential and ADHD mode after depressingly few exchanges before they need to reset. I’m talking like, 5? Just not good enough to really satisfy.
Huge improvements over time of course, but they seem to have plateaud for this particular issue.
I started learn linear algebra and some mid-level math stuff couple of years back. While I really wanted to learn I still struggled to find a good learning path and use a product similar to Duolingo for math.
Eventually I learned some, then stopped using it and then forgotten all of it.
But my point is despite if me really want to learn I found it quite challenging to find learning path that I trusted to be good start for me.
I also need to make the point; I didn’t have 5 hours a day, so this 30-90 minutes in a good day endeavor.
There are still tons of things to solve in this area, how to cross to chasm, a learning path that’s right for you and how can you be sure of that. Learning something in one good 60m instead of 5 different videos and 5 hours. Etc. I think these problems are far from solved.
I also want to get fit but that doesn’t mean I’ll show up in my gym to workout everyday. So even when you really really want to learn you still need to stay motivated.
> While I really wanted to learn I still struggled to find a good learning path
For that 3Blue1Brown is very good for this.
https://www.youtube.com/@3blue1brown
Better Explained as well, but focuses more on how to understand the concept so you can start to learn.
https://betterexplained.com/articles/linear-algebra-guide/
3B1B are primarily math entertainment videos, they don't actually teach you anything.
They are good if you already know a subject well, and want a new perspective on the material though.
You are partly correct.
They have training video series as well. Even one on Linear Algebra. Starts from scratch and works up to advanced.
The issue with sources like these is that they are inherently passive. They don't include exercise sets. While they can be good supplementary material, learning happens by doing, not by passively watching.
I do say this as a fan of both, though.
What you're looking for is not something to help your with your goals, it's something that sells you the idea that you can achieve your goals while seemingly be unwilling to put the necessary effort in.
What if it's simply impossible to learn linear algebra and mid-level math stuff by spending only 30 minutes a day? What if you can't get fit if you don't go to the gym?
It sounds like what they are looking for is a pathway. Khan used to have a really good one with the world of math pathway where you could go for 2 + 2 to Calculus in a set sequence of steps each building on the next. I've tried Khan a few times since they got rid of it and it just feels like you're all over the place and regularly missing skills it expects you to have.
They aren't trying to get fit by not going to the gym they are looking for a progressive overload program for math.
People want to learn at the last moment, when it is needed to know, and then they want to discard that knowledge until it is needed again. We need to implement both in our modern software: it needs to solve complex problems, and it needs to guide a human (or AI) operator in the subtle nuances of that complexity so their solution of the moment is good to last for the time being.
AI isn't that general purpose. If There is a chef course curated / tested with Jamie Oliver I'll buy that over a Claude chat. And by "I" I mean regular folk who are most of the market every SaaS and AI wants to capture.
This is exactly right. I started teaching undergrad CS about 18 months ago after 20 years wandering through other careers, and I quickly realized that students do not need better teaching. They need frequent, direct feedback and an environment that more or less requires them to commit to studying until mastery or to quit early and do something else.
Effective learning techniques have been known for decades, but universities have -zero- incentive to improve teaching quality. There is no standardized assessment infrastructure to compare graduates across schools, or to evaluate the effectiveness of teaching interventions at scale. So any effort put into teaching quality is invisible and thus worthless to a university.
Every attempt to make high-quality instruction available for free has been largely irrelevant to the unwashed masses, because universities have a monopoly on credentialing. The self-starters can learn anything for free, but they were never the problem. It’s the middle 50% that we have utterly failed.
I believe a huge part of the failure is the lack of teaching anywhere, the general idea of how to manage discussions of disagreement. Because that is not discussed or taught, any topic with disagreements triggers emotional reactions in just enough people that no discussions of any progress can be made in such areas.
When the "middle 50%", as you call them, find they need help or really anything beyond what is immediately available they find they cannot articulate their need, nor can anyone around them understand them in their requests. They are propped up with an incomplete script and just left to flounder around, and because they cannot articulate their situation larger society is allowed to ignore them. Which is a completely manufactured situation.
I think learning from another human being live who can understand your confusingly articulated questions and who can adjust their answers for you, another human being, that's great.
I agree, at least in that we need more human contact, not less.
But I will say Gemini / Claude have both been surprisingly good at understanding my half-assed attempts to put a complicated question into a coherent sentence.
Too costly.
Now its viable with AI to do it cheaply.
Try being an English speaker learning Mandarin from Claude
What obstacles do you tend to run into? I'd be inclined to use a Chinese model for that, since Claude was presumably RLHF'ed primarily by native Anglophones.
Have you tried a Chinese-trained LLM for learning Mandarin, and if so, did it help?
>Too costly. Now its viable with AI to do it cheaply.
Where is the data that shows this sort of technology beats normal pedagogy?
From a study I read that came out last year, AI tutors are at best no better than a teacher, and whats worse is that AI tutor use caused many students to become dependent on them. These students could not recall information or perform nearly as well without them.
This is just another example of a long line of behavior from software engineers thinking they are god's gifted people, who can waltz into any industry and revolutionize it overnight with an application.
What do you mean by “learning systems?”
I was referring to systems that are designed to teach people.
SCORM for example.
Scorm is a zip file with html files to structure content any way you want with a minimal amount of mandatory standards.
It was never meant for assessment, nor is it secure. The creator of scorm said to not use it learning had to be secured.
It’s not a system or a framework in any meaningful or standardized sense, and more of a format for compiling content that can be copied between different places.
That's an oversimplification of what SCORM is, but point taken. My comment was about LMS.
It's not an oversimplification of it at all.
I just took the conceptualization of it away from the sufficient experience I have with LMS to do the same with it:
Learning management systems are rarely about learning. Too often, they're first about about managing learning for the learning manager, and teh priorities of the learning manager (maintaining compliance) and that's what the software and sales process aims for.
The learner, actually learning, or being a central focus is a second or third class citizen in such tools.
Yep. Literally every EdTech idea is a rehashing of the same flawed reasoning: we just need better monitoring and metrics to manage everyone’s learning process! No. Learning comes from repetitive deliberate practice in the zone of proximal development. For essentially all knowledge-based skills, what’s missing is the core foundational skill progression that leads to actual mastery.
How many different personal training styles do you see for competitive sports? There’s variation, but only at the margins. Nobody deviates from the essential training core because doing so leads to poor performance and competitive failure. And yet we have a total free-for-all in knowledge disciplines where we just throw untrained PhDs in front of a classroom and assume they’ll produce results. Oh, and we also provide zero tangible incentives for them to improve learning outcomes.
How to teach, and how to learn, and how to learn how to learn are for sure different things.
With post secondaries, one thing I see is the obsession with pedagogy, which makes sense on one hand, where teachers who teach K-12, learn how to teach children, for the unfamiliar - peda is children, and gogy roughly means "how to lead".
If our teaching is rooted actually or inaccurately using a word that is rooted in a practice of how to lead children, it doesn't surprise me that the learning experiences of non-adults, including young adults, are so hard.
I've got to work in K-12, Academia and Industry training/certification, and where they work, they do not share as much as each of the categories would like to think.
Thus, when someone takes a K-12 framework or approach and applies it in industry, it's feasible that it might not work, but the takes on why seem to get ignored.
Fundamentally skills and the ability to use your skills change and evolve and fixes curriculum that takes 1-3 years to change one sentence doesn't really work in more and more places. This is more a fundamental change than a short lived irritant.
They are solving the bloom sigma problem from where I see it.
It’s far from solved.
Qwen 3.6 27B will do that for you, or any other model with a few billion parameters. I don't see any value there other than nice UI on top of an LLM.
You don't have to see the value - the learners have to benefit, being the majority, they usually won't be setting it all up like you are implying on their computers.
Yeah, I think people who really want to learn will very quickly grow out of this.
This is for people who don't want to learn but actually have to for some reasons.
Totally agree that the missing ingredient is making people want to learn. Which seems like something that gets harder every day.
A very cynical way to put it: "Why should I learn something when AI could do for me?"
In a world where most people learn to get a job, its less of a cynical question and more a practical question.
Is it? People want to learn things they think they need. Is there a growing disconnection between what they think they need and what they actually need?
There is a growing lack of motivation.
I mean "People who want to learn will gravitate to learning systems." means that this is the obvious target market for a learning system. Since there are lots of those out there you win on execution.
>If you can find a way to pull in people who don't want to learn, that has value.
Most companies try to address a known market and not create one.
At any rate you have two types of people who don't want to learn:
1. people who don't want to learn and have to anyway, for some reason (generally work related)
2. people who don't want to learn and have nothing forcing them.
Number 1 is a potential market, number 2 seems hopeless.
Most people don’t want to learn, they want a credential. The real hard problem here, that has been sidestepped for decades and decades, is how to gate credentials behind standardized skill assessments rather than branded gatekeepers. Stanford, Harvard and MIT can do whatever they hell they want in the classroom as long as they continue to skim the cream of each generation and no objective assessment exists to demonstrate that they contributed nothing above what their exceptional students probably would’ve done in four years anyway.
As a daily user of Anki and Math Academy, I'm very excited about how software can help us learn.
Now, with AI, we can really build tools that were not possible before. I built for myself AlgoTutor[1], a CLI tool that teaches Go developers algorithmic skills and concurrency.
The recent Understanding floating point[2] from Oz Nova is also a type of tutorial I want to see more often.
[1] https://github.com/zuzuleinen/algotutor
[2] https://proto.csprimer.com/fp/
Massive thank you for the introduction to Math Academy. After a quick look I signed up and it is fantastic - exactly what I wanted.
I'll take a look at AlgoTutor in the future :)
Math Academy is doing it right. Their neat synthesis of spaced repetition and the knowledge DAG is slick, but honestly the extremely fine-grained skill decomposition is what’s missing in virtually all of higher education. I’m trying to follow MA’s example as I put together curricula for my undergraduate CS courses.
I'd never heard of Math Academy. It seems interesting, but quite expensive ($49/month for software which is in beta). Would you say it's significantly superior to something like Khan Academy (which is free) to justify paying that much?
It is indeed expensive, especially considering my current location(Romania).
I have a rule of doing 20 XP per day, and I kept that rule for more than half a year. Maybe the price is also a factor, but I never managed to keep that pace for Khan Academy when I tried it. You can check their current courses here: https://www.mathacademy.com/courses
I think the topics and curriculum are more serious, and I would definitely recommend it if you have kids or are in highschool/uni. I'm 37 now and only do it to keep my brain sharp and build my mathematical intuition and I think is great for that due to its focus on practice and spaced reviews.
I think they have a 1-month trial, and you get your money back if you cancel within 30 days if you want to try it.
I tried Math academy twice and I cannot understand the hype but a lot of people seem to love it so maybe it's just not my style of learning. Each topic seemed to contain too little information to actually grasp it. It was like
Trig 1 -> This is a triangle and trig is about triangles.
Quiz 1: blah
Yes if you can afford it, then it's absolutely worth it. Nothing better at the moment.
This review lays out how you should think about the costs and benefits: https://jonathanwhitmore.com/posts/2024-09-10-mathacademy/in...
It is vastly superior to Khan. Had my kids go through both, so can compare.
That said, on higher levels there's some cruft that trains mechanical skills, and that's on subjects that are needed only in the scope of the US system. For Europeans these should be optional, alas they are not and this adds a lot of tedium and hassle to an otherwise rather frictionless process.
If you spend 30 mins/day studying math then that's $3.30 per hour.
If it improves your learning efficiency by just 10% then it's worth it.
(My son uses math academy 3-5 hours per week.)
I do think Andrew is the right person doing this. Someone mentioned that the return of financial outcome is not good in edTech, and this is exactly why backing by Coursera is a huge advantage. Education-related product is slow, and it needs patience.
The concern, though, is that how to achieve those 3 thesis. Do they train/fine-turn their own model? according to my own experience, the current models are not well trained in personal education context, and it is either too textbook-like or too task-driven.
Yes, Andrew Ng is a great educator, well positioned with Coursera to deliver this, and personalized education (with exercises and feedback) is an excellent potential use of AI.
I guess the real challenge would be initial and continual assessment of each student in order to select the next instruction that they are ready for, but with "LLM as judge" that doesn't seem such a problem, and student feedback can be used to continuously improve these recommendations/selections.
Presumably you would still want to have a human designed curriculum, with the AI being used as the teacher - assessing how each individual student is doing and what they need to be taught next.
I remember reading that Education is a trillion dollar market, so i'm not sure about the financial outcome being poor, only the state of how that money is spent and obligated can differ greatly.
The biggest success that people usually point to in EdTech is Duolingo. They were founded in 2011 and IPO’d for about $6.5b in 2021. A big success by any measure. But that’s the pinnacle. Compare that to DoorDash which is an $80b company, let alone the trillion $ AI labs and you’ll see why the largest VCs hardly touch the space.
EdTech is a broad market.
- K-12, Academia/Colleges/Universities, Industry/Training, Adult learning, increasingly experts teaching directly, and more.
Most people aren't aware of the differentiation of the categories because they get to work inside one for the most part.
The learners in each category are different.
Duolingo had it's share of naysayers and doubters, but it did provide it's own take on a learning outcome.
Your post does make me wonder about the existing spend in Education, and how much could/should be EdTech with today's capabilities, or even the capabilities of 10 years ago, and what looking at that with a lens like yours might show, if anything.
There’s a reason services like GitHub Classroom Replit’s education offerings shut down. Serving educators can be very hands-on. Education can have lots of different requirements (every teacher needs this one little feature), often legal issues (student privacy), and dealing with universities and governments is awful and bureaucratic. Most companies will find it easier to make money elsewhere.
The dirty little secret of EdTech is that there is no money in improving learning because there is no way to capture the value. For all but a handful of highly specialized and insular professions, there is no scalable infrastructure for objectively assessing skill mastery. Most learners don’t want to learn for its own sake, they want a credential that proves their value to potential employers. All of the value in learning accrues to the credential, and everyone in the educational establishment is quite happy reserving that privilege for the accredited universities because they can just bundle more and more products and services into the tuition sticker price without ever having to justify it. Oh, and the student loan racket ensures that enough students are price-insensitive that they’ll never look for alternatives.
The solutions exists for skill mastery.
They're just outside of K-12 and Academia, who don't necessarily keep the gap close to the real world and it's needs post-graduation
The value in learning is what you do with it, the credential can be a part of the journey but not always. People were definitely about the brand, and it's value, of the stamp of a piece of paper, however given the experience of recent graduates the past 5 years, it's revealing itself to not be an aberration that's going to give way to the 500-1000 years preceeding it.
"Conventional wisdom says AI will replace people. I believe the opposite."
So people will replace AI?
No no, the opposite is replenish. As in increase people supplies. As in the Matrix.
Yeah, that was my first reaction to that headline too. Given a second thought, he probably means the opposite of "replace", so in this context, "maintain people." maybe? I know marketing headlines shouldn't be interpreted literally, but it tripped me up at first.
I had the same thought. Then I thought, maybe "replace" is what will become opposite. Antonyms of "replace" are hold keep precede break damage deprive forfeit lose penalize remain take worsen
Hmmm. Maybe they should rethink their motto.
Haha indeed; maybe they just meant "won't replace"
Anti-replace. They are going to breed us.
Speciation. Not extinction.
Yes, it's a bit of an odd h1, website also looks very Claude-generated, but regardless, probably most people understand the intent.
Roughtly, "AI will help people."
Here's I think how.
AI learns skills that people have, and replaces them.
vs.
AI helps people learn better and develop new skills faster.
Same here. What a confusing statement.
I'm just wondering if this is replacing, you know, teachers...
Please, stop destroying human everyday life experience with technology and for your personal wealth.
https://en.wikipedia.org/wiki/The_Fun_They_Had
This is one of weaker Asimov's works, in my opinion, as it merely projects pre-mass education era teaching into the future. Conflating mass education and socialization is pretty myopic when writing about the future with unlimited intelligent robots. You don't have to be taught in one-size-fits-all way to have shared experiences, it just happens to be done that way due to the scale, out of necessity, and many education systems and experiments since the 90's are trying to combine individualized learning with socialization. Unfortunately many post-WWII sci-fi predictions turned out to be self-fulfilling, because a lot of people got conditioned on them and can't imagine breaking the lore which is now a part of their lives.
Care to enlighten the rest of us how this is at all related to the submission? Is it because it's AI it is "destroying human everyday life experience"? Seems this company is trying to find a healthier (and better) way of using AI, unless I misunderstand something?
Not sure I would consider public schools a fun experience though. At least not for the majority of pupils.
The classroom experience cannot be replaced, but obviously Independent learning is still valuable and needed.
If said experience is subpar, e.g. when it happens with 30+ student classes when 80% dgaf, there's no harm in offering an alternative to those who wants it.
And the rest can continue enjoying everyday human life experience.
Is it only me or the website looks vibe coded ? the output feels like a standard claude template.
It is. I just wish people would throw a plain text page with a paragraph instead of this because you bet I am just scrolling through it without reading as soon as Claudism appears
/frontend-design
The home page is entirely written by AI. Riddled with em hyphens. More should be placed on the home page at least..
But really what a wonderful mission. Imagine if they succeed and end up teaching millions and allowing people elevate their home towns/countries. Crazy impact.
[dead]
This looks like another investor play to save Coursera and somehow have a vector (pun intended) to help raise its valuation.
Coursera and Udemy already offer AI learning tools, so why create a separate company and commit $100 million? It looks like the product, commercial relationship, ownership structure, IP rights are still unclear.
Edtech has historically not had amazing venture outcomes compared to saas, ads, rockets, etc.
Given how much people spend on education, there’s no reason AI education software wont be a huge market.
And there’s few people better suited than Andrew Ng to execute this
I’ll give him the benefit of the doubt and hope for the best.
My wife has been a high school math teacher for 25 years and has seen waves of tech luminaries give their visions of how tech will radically improve education and learning. She can’t point to any startup that has had a major impact in improving outcomes. I wish I could point her to something that really will have an impact in, or even out, of the classroom for her students.
That's not unique to tech startups. Almost nothing in terms of explicit methodology has any effect on education outcomes. The literature on education research is of abysmal quality too.
I got to see Luis von Ahn the inventor of CAPTCHA discuss his latest crowd sourced venture, Duolingo, which I'm sure everyone has an opinion about, back in 2013 or 2014 at The LAB Miami in Wynwood.
At that point they had ~24 employees. He said his interest was crowdsourcing and thought most of the engineering effort would go toward the crowdsourced translation, but at that time, it was being handled by just two employees.
They brought on two prominent consultant researchers, leaders in the field of language learning and linguistics. The first question they asked the consultants was which part of speech should they present to the learner first. The consultants didn't have an answer. They decided to dedicate engineers to collecting data in order to answer that question.
A big problem with edTech is that we don't know.
Very cool to hear that you were at the LAB Miami in 2014, Around 2016, I was trying to organize a group of folks to build a variable spaced repitition application that tied to a off the shelf eeg, to figure out the optimal time to learn at one of the local hackathons.. We had a whole team of folks, some of the Flomio hardware people, and other local startup folks. Never was able to build the prototype at a hackathon, but always thought there is a physiological element to learning, and ed tech in general.
> off the shelf eeg
We know that mild stress can narrow focus and help learning. "Neural connections do form when you struggle." [0] A little stress is not only good but necessary for learning while a lot of stress causes a breakdown. The trick is being able to judge the stress level of the student and every person is different.
The idea of using some type of bio metric data from EEG or blood pressure sensors will help determine a little struggle from a lot of struggle in formative assessment.
[0] https://www.youtube.com/watch?v=wh0OS4MrN3E
My understanding (in hindsight) is that success of Duolingo had nothing to do with “crowdsourcing” - is that correct? (Beyond maybe stories for fundraising).
It doesn’t seem to have a lot to say about “ed tech” - popularity of apps seems disconnected to popularity
I don’t understand what you mean by “we don’t know”? Are you saying tech entrepreneurs don’t know education? I agree with that. Or are you saying we don’t know what problem to solve?
My interpretation of “we don’t know” is that the space of possible instruction methods is huge, and we don’t know what works best. It is easy to think of it as a kind of optimization problem, but that line of thinking quickly runs into a bunch of practical obstacles: different people learn differently, studies with good experimental design take time and resources so it is slow to iterate, and good teaching seems to require deep insights into student motivations and incentives, aspects of teaching that are hard to quantify. We also don’t have great ways to measure outcomes; grades are weakly predictive of future success.
The best teachers seem to combine deep thinking about how to teach with intuition and charisma. The ingredients are hard to replicate.
I do think that this is a nut that we will eventually crack, but it is proving to be a very tough nut.
It's also quite possible that the method that works best for one person is completely horrible for another person, but that's just me guessing
There is not a huge market of people who actually want to learn something on their own motivation. Get some piece of paper, yes. So the huge market will only materialize if the degree granting institutions adopt something like this.
Wait, when you say one-to-one you mean AI to Human? That’s not the same, I’m sorry.
May I ask why? If by "not the same" you meant: AI is not at the level of a good human educator, it is a fair argument - but not many humans can afford to have a one-to-one human educator either.
Perhaps I am a bit more optimistic, but if AI could give me, let's say, 90% of an average lecturer's capability, at 10% of the price or even less, that is probably extremely valuable. (For one, I would certainly be interested in learning new things that otherwise wouldn't make sense financially)
Moreover, there are areas of expertise that are very hard to reach without special help. I have a PhD in computer science. Earning that degree was a major time commitment, and it gave me close contact with one of the few experts in a niche area. I absolutely loved the experience.
But I did not learn everything I wanted to know. And I can’t afford to take a time-out to go to grad school again for a second PhD. On the other hand, an LLM, in some sense, has access to the world’s knowledge. With some patience and persistence, you can mine almost-expert-level instruction from one.
When ChatGPT 3.5 came out, I excitedly tried this for a topic that I WAS an expert in. The results were discouraging. The model’s answers seemed to reflect the general misunderstandings that people had about the topic. More recently though, I had to review a paper in the same area, and I used a late-model ChatGPT to help check my work. It was an eye-opening experience because it was no longer confused. And it found longstanding misunderstandings I had, buttressing its answers against my skepticism with citations to original work. I came away very impressed. This kind of AI “rubber duck programming” is my preferred style of use now. I used it just today to help me learn an area of statistics I have always been fuzzy about. This approach definitely requires some careful prompting, but I am optimistic that AI tutors will one day be a real thing. My only worry is that people lose the ability to understand what makes an answer a good one and why we should care about good answers.
>And it found longstanding misunderstandings I had, buttressing its answers against my skepticism with citations to original work.
Did you check those citations to make sure they actually contained the information they were being cited for? Because I've asked difficult questions too and have had mixed results. I would never trust an LLM as a teacher. They make stuff up all the time, and when you're learning you're in the very worst position to be checking the information you're given.
A lot of knowledge in such niches has never been written down. Papers don't contain it. Tacit knowledge, it's in unpublished config files and unpublished experiments that some prof or PhD student remembers, all the things that are not "publishing-worthy" (because not flashy, not novel-sounding enough, or a negative result etc) but crucial for the method to actually work. Most of expertise is like this and this is why spending years in a research group, talking to the researchers, participating in their culture, understanding their thinking over years (or even just months) can be invaluable. There are also many things that people only say after a beer, after you know each other well. Claude and ChatGPT haven't ingested any of this. They ingested the sanitized surface version, the press release / advertisement version. Because papers are basically career marketing and advertisement mostly. The real expertise is guarded and hidden.
This is of course about the PhD level and beyond. For regular university-level education, where the goal is to pass the usual exams, ChatGPT and Claude are more than enough. But for real-world "doing", not (yet).
I just fundamentally believe that AI cannot replace the human aspect of connection. A teacher does more than just teach, they also connect with the students and form relationships. You could argue we may end up with some form of “AI” connection though if that’s a good thing and what is its long term effects on us as a society we are yet to see. Regardless, what I mean to say is I don’t think AI to Human can replace Human to Human as a whole, that’s my point. The marketing copy feels deceptive and not something that speaks to me, I wouldn’t consider that to be same. This feels like it’s marketed towards MOOCs but I am not a fan of education as a product like that either, I come from a place where education is free and this idea of gatekeeping knowledge and forming products around it just doesn’t sit well with me. This just feels like a way to make more money and marketed towards that narrative. You’re right it lowers the cost, but at what cost? MOOCs already didn’t accomplish much the way I see it, I saw much better results from self made paths with things like YouTube or free lectures and/or self made teachers who go above and beyond. That said, I think there is a role for AI in education and I do think it belongs in there, alongside a human educator, in some form.
I didn't have "replace" in my mind. I 100% agree with you that AI cannot replace human educators, but I am hopeful that as a supplementary tool, AI can achieve a lot.
On your point of MOOCs and YouTube lectures: I agree that the latter is sufficient for me, but there's nothing stopping AI from creating something similar, or even better: lectures tailored to my specific interests and needs. MOOC failed for me exactly because it wasn't "for me" - I might as well try and find a YouTube playlist.
These are all tangential, of course. I don't know how this specific thing would turn out.
In some ways an LLM tutor is better. Try getting a human tutor for things like Category Theory or Lean 4.
I believe if you need a human tutor to understand category theory or Lean then you don't have the requiste math/CS background, and a human/LLM tutor won't be able to meaningfully teach you anything.
Note I said "meaningfully" - something with LLMs that deeply concerns me is that they provide edutainment, and people actually think they're learning something. One of the comments on this thread mentioned giving a "dense linear algebra pdf" and having an LLM use the Socratic method. I guarantee they would have learned more if they read the PDF. I strongly doubt they learned anything at all with a bunch of silly "Socratic" questions about linear algebra. But it felt like they learned something!
" if you need a human tutor to understand category theory or Lean then you don't have the requiste math/CS background and a human/LLM tutor won't be able to meaningfully teach you anything."
What an incredibly elitist thing to say.
No, it's an honest thing to say. If you're not able to read a graduate-level textbook and teach yourself category theory then you don't have the mathematical maturity to understand category theory. You need to start smaller: ideally abstract algebra and point-set topology, or set theory and mathematical logic if you just want the CS applications. Likewise with Lean. Of course there is a place for a good human instructor. But mathematical maturity must be developed the hard way.
Even if you are correct, which you aren't, a good tutor almost ALWAYS helps people learn faster. This is just a really odd thing to argue.
I also wish we had an infinite supply of expert teachers available to work 1 on 1 with students at any time day or night for a nominal fee.
Sadly we do not, so something like this will have to serve as a more accessible backup.
Well, hold that thought, what if one-to-one means AI to AI...
I feel like this website would have been better off as just Andrew's letter (halfway down the page) in system default <p> tags.
The AI sloppification is palpable.
Anywho, curious to see what comes out of this! EdTech & AI is a really promising space.
$100M funding is a hell of a lot of money.
Yeah, prominent thing run by famous guy … but why so much funding so early?
What does $100M enable in the next 5-10 years that $25M doesn’t?
Coursera, the company providing the funding, also belongs to Andrew Ng
He's on the board, at least - "belongs to" isn't correct.
From https://en.wikipedia.org/wiki/Coursera
> Coursera was founded in 2012 by Stanford University computer science professors Andrew Ng and Daphne Koller.
founded by and belongs to aren't the same thing
See The Diamond Age: Or, A Young Lady's Illustrated Primer
Website is by Claude (Claude likes beige). I wonder if they'll work on a real product or if this is just a new company to divert money from Coursera.
I'm working on a much smaller idea, but so far it's just me using it for grad school, and a friend using it for his master ciccerone prep. It has been really useful in the class I used it for while building it. Looking forward to playing with more as I start my fall semester(UIUC CS 425, Distributed Systems).
I need to get some model pricing evals in place to see if I can use open weight models to get it cheap enough, since it's just me and not VC money. The real product potential seems to be getting reusable materials like cert prep, continuing education for careers that require it, and the like. But then you're fighting incumbents that fiercely control their moats.
https://studyengine.app/
It's funny that AI even gave us similar themes.
This looks awesome. I'm very excited about how learning will evolve harnessing AI. I'm also a huge fan of Andrew Ng.
I wish they hadn't raised so much money. I'm not sure "large funding round" sets up the right dynamic for what he's trying to do.
> This looks awesome.
I'm curious: what looks awesome? The website was created with Claude. There's no detail about what the actual product/service is and how it will work. Everything is a big idea/promise with no specificity.
I've been working for a few years now as a private tutor teaching proof-based math to adults. Maybe this should give me an incentive to join in on some of the cynical replies this has gotten, but that's not actually how I find myself feeling. I love teaching, especially in the one-on-one format, but it really doesn't scale; not everyone can afford a human tutor and there probably aren't nearly enough of them to meet the possible demand.
If a product like this worked well --- and I agree Andrew Ng's involvement seems like a positive sign --- then that would be great! Something I would wonder about from my past experience using LLM's for technical subjects is how to replicate a particular side of the student-teacher relationship. Sometimes a new student of mine will need to be told that they don't actually understand something as well as they think they do, and that they'd be better served by going more slowly than they might have wanted to. This "talking you down gently" conversation is one I'm happy to have (and sometimes it's actually me who's wrong!) but I'd be interested to see whether and how well an LLM run by a company that really wants to retain customers could pull this off.
It's hard to say without an actual product, of course. I'm looking forward to seeing what they come out with.
100M investment from Coursera? Isn't Andrew Ng a cofounder and still chairman?
To help anyone else who hadn't also previously heard of Andrew
https://en.wikipedia.org/wiki/Andrew_NgThe Digital Maieutic: Socrates and the Art of Prompting - https://news.ycombinator.com/item?id=48342887
Andrew Ng's Coursera courses: https://www.coursera.org/instructor/andrewng
The hardest problem seems less like content generation and more like accurately modeling what a learner actually understands.
Hopefully the app itself won't look AI-generated like the design of the website. This one isn't so bad, but has hallmarks of Claude designs. Nowadays I keep getting shocked to find severely sloppy designs when browsing through Y Combinator startups. For me, this is a signal that the people behind the product don't care much.
So the founder of Coursera is starting a new venture backed by Coursera ? That sounds like the Musk playbook.
I don't really get how this is new or even "AI". Can't you already ask Claude to teach you something ?
"Patiently stays with you until you've mastered new skills", that's so nice. Maybe in the future it will be the other way around, your computer will complain about how slow the user is :)
Reminds me of Karpathy's education startup. I think it had a similar premise. What happened to it?
I think this is a great space because almost any AI educational product can be measured and evaluated by the end user. For example, does the user improve from using your product and do they also believe they do, and do they also LIKE your product? If not, then they have every right to seek out the perfect AI educational product, of which, I expect a variety of.
Personally, I believe the last 40 years or so of academia has failed at creating the proper reactive curriculum for students, so I'm nearly positive anyone that enters the AI education space has a very good chance of proving to their end users that they can out-teach ... I don't know, the entire NY Board of Education, for example.
It is not bold at at all to say K-12 will be eradicated. Higher Ed not so much because many were always buying the status of higher education anyway, so the status in itself still has value as a product.
> It is not bold at at all to say K-12 will be eradicated. Higher Ed not so much because many were always buying the status of higher education anyway, so the status in itself still has value as a product.
It’s funny because I believe the opposite for both of these. Living with a teacher, I’ve learned that the most important objectives of k-12 education are teaching kids how to act like adults in society and behave around others. This is something I don’t think AI will ever be good at teaching.
A few other things, like learning to read and write, basic math, critical thinking are secondary, but important, but, I imagine not the best use case for AI either.
Beyond that however, you’re an adult equipped to take learning into your own hands with all the understanding and agency to make it happen. With the help of AI I don’t think there’s much you couldn’t learn, at least that doesn’t require human interaction (e.g. teaching).
Oh it will be even better than humans some day. Not any time soon, and definitely not when the sort of people training these things range from incredibly awkward around people to downright psychotic in public. But the infinite patience and persistence of coding agents in the form of a teacher for other things than code would be pretty amazing, but we're nowhere close to that.
Case in point, I would love to know who thought deploying a figuratively "demilitarized" sexbot as a teacher in NY town in 2026 was a good idea. And I say that as someone who loved the movie Weird Science. And I'm evil enough to admit that Van Halen's Hot For Teacher was the first thing that slipped into my mind reading the article below. But really? Who?
https://www.npr.org/2026/07/29/g-s1-136072/ai-robot-teacher
Ive actually started trying to build games, starting with Roblox, to eventually build a catalog for this same problem. 1:1 learning in the future is probably going to be 80 - 90% gamified because people like playing games. Spent the last 6 months traveling Asia and everybody was playing games on their phones...everywhere. The first game I released is Math Cadet, basically an adventure game to build up math skills for kids elementary and middle school. I have another one in the chamber thats more science related. Any feedback would be great: https://www.roblox.com/games/129132556457079/Math-Cadet-Adve...
I wonder what they will do different compared to Khanmigo, which wasn‘t particularily successful wirh full-on AI learning assistance. Support for traditional human learning systems?
deeplearning.ai context hub codream.ai OpenWorker LearnVector
This guy never stops.
"We're heads‑down building, and will have products to show by early 2027."
By then frontier models may be able to do this by prompting. Even now it is very easy to get one of the chat interfaces to keep quizzing you and adapt to your level, answer your questions, and in general behave how I imagine a tutor behaves. It lacks integrated stats/progress/gamification/long term memory but those can be added by a simple vibecoded app or are usually unnecessary for actual learning.
I hope they’re planning to build hardware so the learning experience can be close to pen and paper. Keyboards and screens are a distraction and the brain learns better with a pencil than with a keyboard.
I feel the exact opposite. I hand-write very slowly, so having to focus on that instead of the material is literally a distraction, compared to just typing out my thoughts.
The exception to this is if the focus of the learning is the writing in itself. Right now I'm learning kanji, so I need to draw them to focus on and memorize the shapes. That said, it doesn't have to be on paper. Dragging my finger on a tablet or moving a mouse are just as effective as using pen and paper.
Slow is half the point though! Memory has a very limited bandwidth.
The other half: https://www.mdpi.com/2075-1729/15/3/345 and a plethora of others.
>Handwriting activates a broader network of brain regions involved in motor, sensory, and cognitive processing. Typing engages fewer neural circuits, resulting in more passive cognitive engagement.
Yes, that's exactly what I mean. If I'm focusing on what my hands are doing I'm focusing less on the material proper than I'm trying to digest.
The idea is that focus and actually working all of your hand instead of just fingers helps learning the material more efficiently even if your own perception of it does not agree.
Maybe it works for some people, but it certainly doesn't for me. When I'm handwriting, especially in bulk, all I can think about it how much I hate it and how I can get it to end faster. I take fewer notes with more abbreviations that are later harder to decipher.
It's different if it's something like mathematical expressions, because those are never easier to type, but just regular sentences? Absolutely not.
Interesting, I can see where you're coming from. Key detail is the notes are actually not supposed to be read later, just written down - the act of writing by hand helps to commit their content to memory.
I get what you mean, and have noticed that effect on my own too, but the increased recall doesn't supplant the note, it just indexes it. I might hear something and go "oh, I remember making a note about that; it was something about magnets", for example, but I still need the note for the specifics. I haven't noticed any difference between typing and handwriting, in that regard.
There's a huge potential in personalized learning/education space. A big problem in education is that everyone learns different, and teachers / schools have to go by the speed of the lowest common denominator. AI can tailor content to the individual students most optimal way of learning. Whoever cracks that code will literally boost humankind
I have a feeling you've not been around a group of teenagers for a while.
I'm a father of two teenage boys. Both of them could've used "customized learning". What makes you think otherwise ?
2 and 20 aren't really the same at all.
Man, that's a dog of a website.
he has like 20 AI companies now.
Is he a grifter now?
The job postings being all on-site (no remote) is kind of wild.
Not really, have you tried to do anything innovative working remote with a team? In any meeting half the people are checked out watching youtube. The collaborative whiteboarding experience is terrible. Working remote is good for plug and chug type work, but the team dynamic is an empty husk compared to working alongside others in the real world. It can be done if you absolutely have to, but don't try even pretend the effectiveness is anywhere close to being in person.
Yes, I've worked remotely for the last decade+ on many innovative teams and typically find them more engaged, autonomous, and innovative than most of the on-site-only teams I've also occasionally helped.
am i in danger? https://therobotlearns.com/
Mr. Rogers death arbitrages a funny fast film of child education for the modern era.
sounds like mathacademy and many other alike services
in my experience, people are not super interested in using LLMs to help them learn beyond the basic chatgpt chat experience.
“Just one more AI company, I swear…” - last words of an AI-company-junkie.
posted the same day https://pluralistic.net/2026/07/28/hitl-ers/#ai-ai-oh but the TL;DR is that actual understanding of the gaps is required to teach.
Feels like yet-another-ai-product that will be swallowed natively by the big labs.
[dead]
[dead]
[dead]
Ed-tech alumni here. You'd have significantly better outcomes if that 100M investment went towards free breakfast and lunch for students.
That would cover school breakfast and lunches for NYC Public Schools for only 3 weeks. Personally, I'd rather see investment in technologies that can give more than just one large school district a temporary improvement.
Sophisticated personalized education for the masses is finally within reach. Andrew Ng is one of the most significant leaders in ed-tech. I'm grateful that he's pursuing this venture and hopeful for the impact it could have on my family and many others.