I believe a great product in this area is certainly possible and Karpathy probably has the technical chops to create it.
I'm skeptical that the product described in the announcement will result in a business able to generate VC-level returns. Based on my experience as a tech entrepreneur with edu-focused products across multiple startups and decades, scaling sales revenue to significant levels in edu is extremely challenging for unique and non-obvious reasons. None of the edu-specific challenges are directly related to the effectiveness or relative value of the product itself. They are structural problems related to the three main types of buyers, which are: 1. Schools, (K-12, Univ), 2. Companies (employee training, compliance, etc), 3. Individuals (self-learners). There are other types of buyers but they don't represent nearly as much aggregated potential revenue as these three. Note that these are the economic buyers not the product end-users.
Each of the three types presents different challenges from the others, however there recurring meta-themes. The main one is that their purchasing decisions are often driven by factors unrelated to the product. For example, the primary focus of schools is, sadly, not maximizing net student learning. While that is a consideration, it's swamped by other external requirements, many of them wickedly complex. For companies, employee training, whether a regulatory requirement or purely to increase employee effectiveness, isn't core to their actual business - meaning it's a cost not revenue. Individuals on the other hand break into two types, those doing it to actually learn something for themselves, and those seeking certification or external validation like a degree to increase their future employment earnings. Learning lovers are a pure consumer, non-essential, discretionary spending-type sale with all the same well-known LTV and customer acquisition cost challenges. For certification/validation buyers, they aren't primarily focused on the learning but rather checking a particular box that has high perceived value to external parties.
While I agree that creating a new large and growing type of customer in addition to these three is possible, it's a huge lift that's risky, expensive and will require substantial sustained investment over time - all ahead of meaningful revenue. So, I wish Andrej et al the best and I'm rooting for them but I'm not optimistic they'll build a high-growth, high-revenue sustainable edu business around AI as the primary value differentiation.
It’s true that these markets are the huge ones today.
But there is another buyer that your post leaves out: parents. This is not the same as individuals/self-learners (the buyer is not the user, exactly) — but the buyer has strong incentives that align with student performance that are different than the others.
Historically, this has been a very challenging market to scale up well — exceptional 1:1 tutors are rare, and hard to train, they’ll be expensive and often form a relationship with their tutees that incentivizes disintermediation of any platform that tries to capture some fraction of the value by offering a two-sided marketplace (YC has funded at least one of these!)
However, AI has the potential to dominate in this market. Most tutors are actually not amazing, but 1:1 tutoring is just so much more effective than basically any other kind of educational context (two-sigma improvements!) that it doesn’t super matter. AI could be a much more effective tutor — at a cost substantially lower than a human tutor. Of course, wealthy families may still want to hire the well-pedigreed, role-model-level, truly exceptional tutors. But folks who are lukewarm about their private (or small-group) tutors may just give the AI tutors a shot—and find that they’re better than what they were paying for.
Or not! That’s what makes it a risky bet. :)
But it’s a large, $50B+ market today, and growing fast — not one to be discounted.
> but 1:1 tutoring is just so much more effective than basically any other kind of educational context
Do we know how much of that effect is the "personalization " of 1:1 learning (something AI can replicate) vs the effect of interacting with another human (which AI might not be able to replicate)?
The studies almost always compare 1:1 learning with learning in groups, comparing two different modes of interaction with another human,
But it’s reasonable to ask whether an AI can (today?) be as effective as a human at delivering (or even identifying, generating) that personalized interaction.
> But there is another buyer that your post leaves out: parents.
I did consider adding a section to my post breaking down all the smaller sub-categories of the main three, of which "Parents" is a significant one.
> Historically, this has been a very challenging market to scale up well
Yes! And for several different reasons. Labor costs, which you mention, can be a major one for products which offer highly individualized interaction. I agree that LLMs should be able to substantially disrupt existing tutoring type businesses. Another major challenge of parent-direct sales is that they aren't a well-aggregated market. Scaling sales there gets costly in average cost of new customer acquisition for new brands. It's basically a direct to consumer business, with all the well-known challenges that presents to fledgling starts. Of course, there are examples of companies that managed to do it, but as you say, it's risky.
I agree with all the shortcomings you mention with existing tutoring and the potential of AI here seems obviously compelling. In fact, both the technologist and the passionate believer in learning inside of me are giddily enthusiastic about the vision you paint. However, as a serial tech startup entrepreneur who recently retired (somewhat) early after a multi-decade, multi-startup career that was (eventually) quite successful... I'm also, sadly, now too jaded and battle-hardened for my edu-idealistic enthusiasm for that vision to survive for long.
Assuming you live in urban or suburban North America, to see what I mean, you need only take a trip to your local "learning center" franchise location between 3p-6p. Park yourself near enough to the entrance to observe the parents dropping off and picking up their middle and high-schoolers in waves at the top and bottom of every hour. You'll probably also find several parents waiting in their cars. Be bold and chat with few. Ask if they've considered or tried computer-based tutoring (either Zoom-style or automated). I think you'll find many have. Now ask what this tutoring center offers that's better enough to be worth the increased cost and hassle. If you're good at customer interviewing (and active listening), you may be surprised at what you learn.
As the parent of a 9th grader who's recently paid for both learning centers and private tutors, I can tell you the thoughtful parents will share that an in-person human tutor offers a few key things an automated (whether AI or not) or even remote video tutor doesn't. For one, a local tutor, whether at a center or private, is probably deeply familiar with what the local school's curriculum is in the relevant subjects. While schools are all required to teach to state and federal guidelines, the texts used and the sequencing of the instruction and granular testing varies from district to district (and even school to school) more than you'd think. The tutors at our local center knew things down to the level of individual teachers at each of the two local high schools and they used this knowledge to customize the objectives week by week.
If you're not the parent of a 9th grader who's falling behind in math, you may just think "The student can tell the AI what their current curriculum module is." If you ARE such a parent, you already know why that is an unreliable solution! For the rest of you, understand that there are two broad types of students getting tutoring. One is a highly-motivated self-learner who gets good grades and whose parents are paying more just to pile even more icing on that already well-iced cake. The other kind is more like my kid who, unfortunately, takes after Dad as far as school goes. A non-motivated, highly-distracted learner who is extremely bright and intellectually capable but simply does the bare minimum in school to get by. For us, the parents of this wonderful kid who shines so bright in every way except academics, the learning center offers something an AI can't, a real person whose presence enforces actual engagement. Another subtle yet even more important benefit is an environment full of peers who are all exhibiting that alien "learning behavior" together. For a teen struggling with even engaging in academics, this peer modeling is invaluable.
So, yes. AI has the potential to be revolutionary for student tutoring - in concept. And it will be revolutionary for some students. Yet somehow, in practice, this amazing solution which should simply replace human tutoring - won't. And the edutech entrepreneurs on the front lines trying to figure out why will discover there lurks a subtly wicked complexity under this apparently highly-tractable problem. And these thorny barriers to adoption are all hidden behind human psychology and teen behavioral dynamics. Thus, my OP was really intended as a cautionary tale and warning to all startup entrepreneurs ready to set sail in uncharted edutech waters that "here be dragons!"
Agree with everything you say here. I don’t have a 9th grader (yet!) but I tutored for years. The interpersonal component of tutoring absolutely dominates over the content component.
I think the “this time it’s different” argument is grounded not in “better content delivery” or “cheaper than a person”—which I agree are 100% not winning strategies.
Rather, the argument is some form of “so brilliantly delivered, so charmingly interpersonal, made so relevant to personal interests that even the most jaded, disinterested, burnt out 9th grader can’t help but learn the material”.
We’re not there yet, but even old-school automated tutors were fine at content and cheaper than a person. What modern AI brings is all the rest of what makes 1:1 tutoring great.
That’s what Karpathy et al should be building, IMHO!
I agree that knowledge of local schools/teachers is helpful. But this misses the forest for the trees. My point (as another commenter who asked about parents as a market elsewhere) is that this could upend the entire education system. Who cares what my local school teaches for biology, if what I really need is for my kid to learn bio and get a good score on the AP test?
Furthermore, the fact that some people use in-person tutors doesn't mean that no one uses remote tutors. This absolutely happens. We live near Stanford, and there are tons of Stanford students who advertise their tutoring services on ND. These kids didn't go to local schools, and they don't know the local curriculum — they're from all over. But parents pay to have smart students tutor their kids.
I imagine that these parents would also pay to have an AI bot tutor their students if the results were good, the price were favorable, and the bot were (of course) available 24/7.
BTW, I'm curious where you live that "education technology" is called "edutech". You mention having been in the field a long time, but I've never heard it called this (always "edtech"). Where have I not been that it has this alternate moniker?
To me personally, this is the best comment I've read in my time on HN since 2011 or so. I am also building a learning platform as potential lifestyle business, having been a user of such platforms and worked in large corporate I fully understand that the groups you identify are absolutely true. One important point is that you can make successful business helping people prepare for the certificates without being the issuing party - not VC level, but still. This way you can cover some of the checkbox checkers and self-motivated learners at the same time.
Self motivated learners who will be willing to pay you can also be groupped in a few groups - the ones seeking future employment in tech sector, looking to understand the technology better and showcase it on their resume with some projects that your platform will allow them to complete and ones doing this purely for fun or to keep up, I believe the latter group is very small.
Platform alone, even with tons of learning materials will not be enough to get this off the ground, you will either need to leverage your existing audience, bringing us back to the fact that everything is an attention economy now, or, you'll need to be very advertisement heavy on YouTube, which you can bruteforce with VC investment, see skillshare for example. NordVPN took off the same bruteforce way, although in a different product category.
My platform will combine technical learning with in-material challenges, sort of a sprinkled leetcode if you will. The idea to implement a LLM side-tutor also crossed my mind.
I was thinking of mentioning that I believe a lifestyle business in edutech is totally doable but my post was getting pretty long. Assuming much more modest requirements for revenue and growth, I'm actually fairly bullish on this scenario. The one proviso is that, in my experience, you need to structure the new business to survive a (potentially) very long period before reaching break-even. K-12 sales cycles can be glacially slow to close full district deals. If you can structure your product and pricing model to also generate some faster (though smaller) revenue from individuals, parents, private schools and single-classroom teacher-directed funds this can help provide some partial life support during the long winter before your first harvest.
Assuming you're able to (or maybe even enjoy) optimizing a business to operate for a few years on hand-to-mouth revenue and can get to product-market fit with few iterations then the AI buzz will make for an especially good entry window for new product offerings in edu. Most of your customers are already aware of AI and open to the possibility it may change things for the better. This should make getting classroom pilots easier.
The final thing I'll share is something I'm sure you already know (but I'll mention for others). It's pretty rare for long-term, lifestyle-scale, edutech businesses to get "good" exits by normal tech startup standards. Think around 1x to 3x trailing twelve month revenues with the higher end of that range only for those having proven consistent multi-year profitability.
Thanks for saying the quiet part out loud here -- everyone wants to believe in education improvement but there's a harsh reality of the industry that is pretty unpleasant to confront.
I think the only hope would be to run a pilot and show such an improvement in student testing levels that it would be fast-tracked or mandated for approval.
…or be accused of cooking the books, labeled as anti-union, or other ad hominem attacks.
I’m not a fan of SV edutech ventures for various reasons, but the school districts that they hope to sink their pointy teeth into have very effective immune systems against change.
Yes, I LOLed at your comment. K-12 is especially soul crushing for tech startups, which is why I now think of K-12 school district buyers like Obi Wan thinks of Mos Eisley.
Maybe part of it is seeing education as a massive untapped wallet? There's a reason Teachers Pay Teachers exists and frankly, given the greed of Silicon Valley and existing ED companies, isn't it an environmentally created immune system?
Teachers and administrators have decades of experience at getting treated as targets for culture-war grandstanding. Having a 28yo tech CEO barely older than their students tell them that a cool website will reverse the effects of poverty is probably just a different flavor.
> So, I wish Andrej et al the best and I'm rooting for them but I'm not optimistic they'll build a high-growth, high-revenue sustainable edu business around AI as the primary value differentiation.
I don’t have an X account and as such might miss additional information from subsequent replies but there is nowhere in the post mention about building a high growth/high revenue business? Maybe Andrej just enjoy teaching and sharing knowledge and want to bring like minded, well off people together rather than creating an unicorn?
Or do you mean by your post that it is not (or is very hard) to even reach non negative revenue to keep it sustainable on its own?
I considered including an aside mentioning I think lifestyle-type edutech startups are entirely doable, especially for those who can survive long periods before meaningful revenue without outside investors beyond friends/family (which probably includes Andrej). I posted a related reply here: https://news.ycombinator.com/item?id=40987394
no, GP just assumes "high-growth, high-revenue" business is the goal, when in reality Andrej is probably worth 9 figures already and has a bigger mission than making "VC-level returns"
Their goal might not even be to build a high growth high revenue business. Sustainable yes. So perhaps low growth, low - med revenue and who knows potentially high revenue.
If anyone else had done this, I'd be very skeptical but give Andrej's penchant for teaching and how good he is really at it, I am super excited about this endeavour.
Let's not forget the inbound and organic interest this would drive. For the intro LLM Course, people are already submitting PRs and Andrej is requesting them not to. Imagine that!
Update June 25. To clarify, the course will take some time to build. There is no specific timeline. Thank you for your interest but please do not submit Issues/PRs.
I think a lot of group courses/recorded courses don't work/lead people to give up because the context provided is not intuitive to everyone's mind and requires significant googling (or ChatGPT chats) for a concept to click before moving on.
Having that ability to pause and clarify with for as long as needed (and have it dumbed down as much as needed) seems like a step in the right direction.
Yes. I was really happy, when I discovered GPT3, because I could ask infinit „but why“s without being worried about annoying the tutor. Would be great, if the LLMs wouldn’t be such push-overs. I’d like to hear more „you’re wrong!“.
I am the founder of an industrial edtech company, with some prior experience in the K12 space.
Your analysis of the market structure and the challenges is consistent with my experience - I'd like to add 2 points though:
1. Building on what you've said about VC-level returns, while edtech (and industrial edtech) may not necessarily be fields where VC funded unicorns can actually form and then survive to profitability, I believe there is sufficient need and space for companies that are investable and can grow into profits / valuations in the 10s /100s of millions - they can follow the startup format, etc.; just not VC investable, as you've noted.
2. In the space of skill development (i.e., adults / professional learning), I would like to add that for many skills, as industries get more sophisticated (I am taking of manufacturing, automotive, energy, oil & gas, infrastructure, petrochemicals, infrastructure), there is value for the company in upskilling their staff - simply because that does directly convert to profitability if done well; and at least puts a floor on issues that poor or absent training may cause - like safety, quality, etc. So even if it is a cost, it is a cost that can directly affect their topline / bottomline, and I believe there is an opportunity here.
I think bootstrapped with a small team of under 10 people, Karpathy can get to revenue of $20M just from the developer market. That's a pretty great outcome if VCs aren't involved.
Great observations. I've been in the edu (coding school) space for over a decade and it tracks with my experience (though it seems you've tackled a lot more market segments than I have). I'm very bullish on AI + Education. If Karpathy's announcement doesn't track, what do you think might work instead?
That's a good question to which, unfortunately, I don't have a good answer. None of my startups managed to make our edu-focused products long-term successful. Fortunately, we designed our major technology stacks so each could result in products in at least two different markets. While our edu products never did better than break-even for us, our non-edu products usually did much better with several becoming breakout hits. Personally, I'm very passionate about the need for innovation in education, likely due to challenges I had with traditional school. But after spending many years and substantial funds, I eventually stopped trying to do edu-focused spins of my products (despite clear potential of the underlying tech). Although I now understand many of the reasons the edu market is so hard and counter-intuitive, I also learned succeeding there requires vastly different skills and interests than I have (or want to have). K-12 was especially frustrating for me because success there is so untethered from actual product effectiveness, along with being unbelievably bureaucratic, largely opaque, and having a glacially slow sales cycle. As someone who started out so idealistic about making education better, actually trying to build a successful tech startup selling to K-12 can be soul-crushing.
I agree that in terms of increasing actual learning, AI could be revolutionary. I commented here because my experience was specifically being a startup trying to sell products based on revolutionary new disruptive technologies into edu.
I'm B2C and selling ed to students is an eye-opening experience. It's actually very easy to get people to pay, so the problem isn't conversions. The problem for me is that if you keep following the money, it's natural to become predatory because the most vulnerable need the most education. The more vulnerable, the more likely they'll buy hype. You're essentially selling to the most needy, which feels like payday loans and that sort of business. I don't have it in me to deploy the sales/marketing playbook on vulnerable audiences.
So, the advice of selling to businesses seems better, but I also hear you on selling to K-12 or higher ed as having all the negatives of enterprise sales without the lucrative upside.
So then we're left to selling ed to businesses. But I feel despite all the lip service paid to training and upskilling, corporations don't *actually* want to upskill their workforce. Or at least not at the sacrifice of productivity.
Education is very unnatural, but that's what makes it a worthwhile problem to solve, too.
There's a difference between the educational-industrial complex (or the ed landscape in terms of market), and motivated learner-centric models which can be done independently.
As much as degrees and certs are important in the current paradigm, it's long been the case in tech that credentials matter less than output, skills, and soft skills.
As much as I personally believe in a stronger credential system, the shift is toward individual learning. The popularity of coding bootcamps in the tech space I think demonstrates that.
Personal AI tutor? Without the pesky SME credentialing? Sign me up!
IMO, a lot of that leads to sink or swim. Not that elite institutions are perfect (by a long shot). But largely cater to kids (and families) that can do things on their own with help from qualified people and you almost certainly get better results--and lower costs--for them.
> They are structural problems related to the three main types of buyers, which are: 1. Schools, (K-12, Univ), 2. Companies (employee training, compliance, etc), 3. Individuals (self-learners). There are other types of buyers but they don't represent nearly as much aggregated potential revenue as these three.
What about parents? That would seem like a natural market for a product that could replace expensive private school tuition (or paying high property taxes in order to be in a good public school district). I recently saw a literacy product aimed at parents, and which costs $50/month. They justify their pricing based on its substitutes being a tutor or private school tuition. While this is a bit of a stretch for a literacy tool, it would be a very fair comparison for a very good AI tutor that teaches the child and has a dashboard/summary for parents.
> What about parents? That would seem like a natural market for a product that could replace expensive private school tuition (or paying high property taxes in order to be in a good public school district)
Quality of education is one factor among many when making the choice to send your child to a private school or move to a prosperous community with well funded public schools.
Another major one is social clustering, premised on the expectations that children will pick up beneficial cultural traits from similarly higher privilege peers that will help them advance.
Edu software can't provide the social benefits of exclusive schools (whether they are private or public). Parents know that which is why if they can afford to, they avoid underperforming schools.
> or move to a prosperous community with well funded public schools
To be clear, in most highly advanced democracies, public schools are mostly nationally funded, not locally funded. The US is a rare outlier, where public schools are mostly locally funded. Most people from both sides are surprised to learn about the other: people from US to learn about "the majority" and non-US people to learn about US system. In most places that use national funding for public schools, if you wish to select a school system with better funding, you pay for private school.
> Edu software can't provide the social benefits of exclusive schools (whether they are private or public). Parents know that which is why if they can afford to, they avoid underperforming schools.
Well, the underperforming schools also have worse teachers, and can be dangerous. It's not like the reason people avoid bad schools is because of the software. A parent recently told me he's sending his son to a private middle school because he's worried about gangs and bullying at the public school. An AI chatbot wouldn't bring those risks (though it also wouldn't deliver benefits of social interaction with well-behaved, bright children, either).
>I recently saw a literacy product aimed at parents, and which costs $50/month. They justify their pricing based on its substitutes being a tutor or private school tuition.
There's a software startup aimed at teaching 3-5 year olds reading called Mentava that charges parents $500/month (not a typo). And there are plenty of online tutoring services for kids at the $200/month level.
I think educational software aimed at parents has traditionally priced themselves too low, hampering the development of what's possible.
Bingo. I expect that the "home schooling" market will balloon in coming decades, and that it will no longer be called "home schooling". It will be "independent schooling" or some such thing, and will be extremely popular. Combine this with remote work, and all of a sudden parents can live in all sorts of places that would have previously been impractical or impossible.
Coursera seems to have made a (reasonably) successful business out of it. $1B market cap and almost $200M revenue. Now of course it is not the scale of other SaaS/Enterprise/Social Media outcomes but there is a market here, I think it's all about product vision and creation of new credentialing systems / disrupting the way skills are assessed at a fundamental level. There is clearly a problem with the cost/value ratio with traditional 4 year universities. I also agree selling into the same system is hard.
I do wonder if recent (Covid onwards) trends toward increased home schooling blurs the line between 1 and 3, or perhaps manifests a new segment that could be meaningful; homeschool pods seem to be quite outcome oriented and don’t have any of the bureaucracy of schools. Also I assume for now they are non-material from a TAM perspective, but interested if you have any thoughts on that trend.
Didn't Apple make a great deal of their revenue in the early days (1976-1985) by selling to schools? I seem to recall that schools were their main focus. A $3000 Apple II in 1980 would be $12,000 in today's money, and schools often bought many units. Is Karpathy's product really that different from an Apple Computer in how it would be marketed to schools?
Apple gave schools steep discounts, with the idea being that young people learning on Apple computers would have parents purchasing computers their kids were familiar with. Later, those kids would become adults purchasing the computers they grew up on.
Let’s view it’s a different kind of investment, like an investment in a child. The return is a bright future for humanity when you educate.
Having a scalable technology that can teach and reinforce learning for humans will change the world. The human potential of teachers will be freed and human intelligence could rise all minds.
No, the potential for this is large. For, schools and universities just $10-$100 per student per semester is enough, and cheap in comparison to teachers when they can reduce teaching and lecturing hours.
If it works. Nothing similar in "ed-tech" has actually offered value thus far.
Excellent analysis. I know you are not in the domain of health care, but would you be able to give a similar analysis for health care tech? And likewise anybody else?
Sorry, I've never done health tech. Based on my limited external observation, it appears to perhaps have three roughly similar types of buyers with health plans/hospitals subbing for K-12/Univ along with many of the same wicked complexities, such as slow sales cycles on the large institution side and high costs of customer acquisition on the direct-to-patient side (basically a consumer biz). Worse, I imagine the regulatory requirements get even more byzantine and opaque in healthcare than edu, so that's another reason I never considered going anywhere near healthcare tech as a startup entrepreneur.
Yet some people make it work and my hat's off to them. They are far braver (or more foolhardy) than I!
I'm skeptical that the product described in the announcement will result in a business able to generate VC-level returns. Based on my experience as a tech entrepreneur with edu-focused products across multiple startups and decades, scaling sales revenue to significant levels in edu is extremely challenging for unique and non-obvious reasons. None of the edu-specific challenges are directly related to the effectiveness or relative value of the product itself. They are structural problems related to the three main types of buyers, which are: 1. Schools, (K-12, Univ), 2. Companies (employee training, compliance, etc), 3. Individuals (self-learners). There are other types of buyers but they don't represent nearly as much aggregated potential revenue as these three. Note that these are the economic buyers not the product end-users.
Each of the three types presents different challenges from the others, however there recurring meta-themes. The main one is that their purchasing decisions are often driven by factors unrelated to the product. For example, the primary focus of schools is, sadly, not maximizing net student learning. While that is a consideration, it's swamped by other external requirements, many of them wickedly complex. For companies, employee training, whether a regulatory requirement or purely to increase employee effectiveness, isn't core to their actual business - meaning it's a cost not revenue. Individuals on the other hand break into two types, those doing it to actually learn something for themselves, and those seeking certification or external validation like a degree to increase their future employment earnings. Learning lovers are a pure consumer, non-essential, discretionary spending-type sale with all the same well-known LTV and customer acquisition cost challenges. For certification/validation buyers, they aren't primarily focused on the learning but rather checking a particular box that has high perceived value to external parties.
While I agree that creating a new large and growing type of customer in addition to these three is possible, it's a huge lift that's risky, expensive and will require substantial sustained investment over time - all ahead of meaningful revenue. So, I wish Andrej et al the best and I'm rooting for them but I'm not optimistic they'll build a high-growth, high-revenue sustainable edu business around AI as the primary value differentiation.