Week in Edtech 3/12/2025: U.S. Department of Education Massive Layoffs, AI in Schools, Cyber Attacks on K-12, Stanford’s AI Policy Shift, Duolingo’s Playbook, AI-Driven Procurement, Sanoma’s Global EdTech Expansion, and More!

This Week in Edtech, Ben Kornell and guest co-host Matthew Rascoff, Vice Provost for Digital Education at Stanford University, break down the latest headlines shaping education, technology, and policy. From major federal education cuts to AI’s growing role in schools and higher ed, this episode unpacks the forces reshaping learning today.

✨ Episode Highlights:

[00:06:02] U.S. Department of Education lays off 30% of staff.
[00:12:07] Education research funding faces uncertainty.
[00:17:26] 80% of teachers now use AI in lesson planning.
[00:18:57] 82% of K-12 schools report cyber incidents.
[00:21:26] Stanford’s new policy bans banning AI in student work
[00:28:07] AI’s future in education—Matthew Rascoff’s “Humanity in the Loop” framework.
[00:36:11] Duolingo shares its company playbook.
[00:42:12] AI-driven public education data and procurement.
[00:44:11] Sanoma’s latest acquisition in AI-powered learning.

😎 Stay updated with Edtech Insiders! 

• Follow our Podcast on:
  • Apple Podcasts
  • Spotify
  • Amazon Music
• Sign up for the Edtech Insiders newsletter.
• Follow Edtech Insiders on LinkedIn!

🎉 Presenting Sponsor:

This season of Edtech Insiders is once again brought to you by Tuck Advisors, the M&A firm for EdTech companies. Run by serial entrepreneurs with over 25 years of experience founding, investing in, and selling companies, Tuck believes you deserve M&A advisors who work as hard as you do.

Transcript

[00:00:00] Alex Sarlin: Welcome to ed tech insiders, the top podcast covering the education technology industry and funding rounds to impact AI developments across early childhood, K 12, higher ed, and work. You'll 

[00:00:17] Ben Kornell: find it all here at ed tech insiders. Remember to subscribe to the pod, check out our newsletter and also our event calendar.

And to go deeper, check out ed tech insiders. Where you can get premium content, access to our WhatsApp channel, early access to events, and back channel insights from Alex and Ben. Hope you enjoyed today's pod.

Hello EdTech Insider listeners. I am so excited today to bring you another edition of the Week in EdTech with Matthew Raskoff, Vice Provost for Digital Education at Stanford University. Matthew's been a longtime friend of EdTech Insiders. He's been on the pod before, and today he's our co host. Welcome to the pod, Matthew.

[00:01:04] Matthew Rascoff: Thank you so much, Ben. So glad to be with you. Sorry to be missing Alex, but, but I'm excited to engage with you and with your listeners as well.

[00:01:13] Ben Kornell: Yeah, and for those of you who are just tuning in, Alex welcomed Anders Sarlin to the world just a month and a half ago. So, uh, this is part of our series of guest hosts. And Matthew, you know, in terms of your journey, you've been leading digital transformation in higher ed across the country. You were at Duke before, and now you're at Stanford.

Just tell us a little bit about what your work is.

[00:01:37] Matthew Rascoff: Yeah. So, um, I've been at Stanford for four years. Um, I lead this office called Stanford Digital Education. And we do mission driven digital learning and our thesis is that digital transformation on its own is not enough. It has to be connected to the educational and academic mission of institutions more directly than it has been elsewhere.

And it can't just be driven by revenues. It's got to be driven by core values and beliefs that tie into like what we do here at Stanford. That's the way to build. You know, online learning offerings that are truer to what we have that are authentic. But I think it's also what the world needs from our institutions.

And, and I think, you know, there's two separate conversations that are sometimes happening at institutions. One is about this kind of, you know, educational, you know, impact, access, equity, those questions. And then on the other side, there's the digital transformation questions like in CIO offices, sometimes IT questions.

And, and our effort is to bring those together and to say we need to put all the tools that we have at our disposal, the technical ones and the human ones, in order to truly make education more accessible, more affordable, more democratic.

[00:02:48] Ben Kornell: It's so inspiring. Every time I hear you talking about it, and I, as a alumnus of Stanford, I also see so many ways in which you all are walking the walk. And so, um, you know, great to have you on the pod today, but those of you who need to check this out, um, will include some links in the show notes.

Matthew's work has been really incredible. Before we dive into this week's news, just to give you an update on EdTech Insider stuff and the pod, we've got a couple great events coming up. March 24th and 25th, we are co sponsoring the Common Sense Summit. On kids and families. You can check out our newsletter for a discount ed tech 10.

And you can go to that conference last year. We met Hillary Clinton there. Sam Altman there, uh, had some great interviews with Sal Khan. Um, so it is really a who's who of people who care about child development. Second, we have our Designing Education Games in the Age of AI webinar that's on Thursday, March 27th from 12 to 2 p.

m. We've got leaders from Minecraft, from Roblox, Legends of Learning, as well as some of the most up and coming AI startups. This is really around the intersection of games and learning. And then to top it all off, we have the mother of all happy hours, Monday, April 7th, from five to seven 30 at ASU GSV.

It's the, um, ed tech insiders, happy hour. So grateful to have Google as our big sponsor for this. We've got started EDT and partners, magic ed tech pen GSE, as well as cooling and tuck advisors coming together. This one will have about 400 people. So make sure you sign up. Now we've got our waterfront location reserved.

All right. So. We're going to see you at one of those events. I'm certain of it. But if we don't on the podcast just this week, we've launched Michelle Rhee from EO ventures. You'll remember Michelle as the chancellor of DC public schools and Lawton Smith from literal a long, long time at tech insider friend and literacy advocate.

So much going on in the space. You know, Matthew, as we sit, sat down, normally we start with A. I. And tech and we were just having a little bit of a conversation. It's like the drum roll of A. I. News. It almost is is we're losing the signal for the noise. But everybody right now is talking about what's going on at the federal government level.

The most recent news being the layoffs of Yeah. From what I can tell, a third of the staff of the U. S. Department of Education, you know, the headlines say 50%. But then when I do the math, I'm like, wait, this is 33%. Um, but it seems like there was, uh, you know, on on Tuesday, there was a memo sent out that Offices would would be closing.

It would be work from home. And then a number of people were given the pink slip. This is in addition to 300 that took an early retirement or early severance. And so the kind of doge, um, effort has really spun over to the U. S. D. O. E. So for this and just overall the kind of federal role in education. What are you?

How do you make sense of this?

[00:06:02] Matthew Rascoff: There's a framework that has gotten traction at Stanford. It's around this idea of an academic social contract. And it's about the relationship between government and educational institutions that has been renegotiated several times in history. And I think this is one of those times when that contract is being renegotiated.

You know, what's worked in the past may not be working for us in the future. It's clearly, you know, we're getting that signal. You know, these agencies are staffed the way that funding models are changing. Um, the, you know, the agencies that flow research funds like IES at, in the education department were already cut a couple of weeks ago.

So it's, it's clear that like, you know, you can't have a contract unless you have two sides working together. And I think this is true of K 12 schools and higher ed. Like, like the government is now. Kind of changing its position on what it wants from us. And we're going to need to figure out what, what that return offering is going to be.

So if there's going to be less basic research that's funded by the government, then it's going to have to change. And like the, the basic researchers are going to have to adapt, either find new patrons to support it, or we'll provide less of it. And I think the same is now coming for the education department.

I mean, it's not yet clear to me where those cuts were within the department. It looks like the office of civil rights. Was gonna be moved over to justice. I saw that the federal, the, the student loan program may be moving over to treasury. So it's not yet clear like which of those things are gonna be provided by other agencies and which of those might not be provided at all, as seems to have been the case with IES, it's, it's a little bit hard to know like what, what the indicator is, but I think this framework of like a negotiation that's two sides.

It helps us think in education about like, okay, well, like if, if they don't want these things and what do they want, if we're not going to have, you know, as significant investments in this one area, where will those investments be going to, or maybe there, they won't be coming at all. And we'll have to pivot.

[00:07:58] Ben Kornell: Yeah, I mean, what I love about the framework is that it doesn't assume a passive. Okay, we've just got to take whatever happens that it really is a dialogue. And when you know, it's, it's us and them. It's actually we, um, there's a degree to which, you know, Alex and I've been having the conversation. What is education for?

And how, you know, how do we wind that up? And there's probably been no greater critics. Yeah. of, you know, what's been going on with FAFSA, with some of the student loan programs than us on the podcast. But this, you know, I think the, the real challenge in this negotiation right now is there's just, um, there's information asymmetry.

We really don't know what's permanent here. What's temporary. And where other pieces are moving. The rumor I've been hearing is that they're talking about combining Department of Labor with Department of Education, which, by the way, depending on your political proclivity, that's a pretty interesting idea of, like, actually lining up a talent pipeline from cradle to through career and and so forth.

And yet, I think there's, right now, we're just seeing a lot of people kind of curled up in the fetal position saying, Oh my God, this is happening to us. So I, I love the active, active role you're talking

[00:09:20] Matthew Rascoff: Yeah. I was, I taught my last class at the GSB yesterday with my co instructor, Rob Erstein. And you know, we were, you know, talking about this with the students, like this is clearly a crisis. The status quo is being detonated in front of our eyes, but education is still going to be here. You know, it's going to look different.

It's going to be funded differently, but it's not going away. And I think, you know, those who figure out how to pick up the pieces. And, you know, put them back together again, possibly in some new configuration that might actually serve learners more effectively. They, they may emerge out of this, not unscathed, but, but possibly successful.

So like that, that, that to me feels like the moment we're in, it feels a little bit like the pandemic, just a little bit like the financial crisis. Um, 11. I don't know. I don't know what your methodology is. Like those, those are the three crises that I've

[00:10:12] Ben Kornell: Yeah, where,

[00:10:13] Matthew Rascoff: career.

[00:10:13] Ben Kornell: yeah, where your fundamental truths are, like, questioned or put, um, put up for grabs. I mean, I, I will say for K 12, the reaction has been more muted. In K 12, most of, and for those of you who aren't from the U. S., like, welcome to, you know, U. S. education funding. There's so much more that's controlled at the state and local level that for K 12, we're really talking about title funds and the disbursement of those funds, whether it comes from the U.

S. Department of Education or the Treasury or whatever, really doesn't matter as much. I think the K 12, um, Implications really have to do with research grants and just funding for K 12, um, evidence and efficacy. And that's where we're seeing, uh, you know, a bunch of the social service, uh, grants that were really around what's working for kids.

That's kind of the jump ball in the air. Whereas, like, core universities, which are delivering the services, I think there's real questions around how will FAFSA work? How will student loan programs work? And We've got probably more questions than answers at this point.

[00:11:21] Matthew Rascoff: Accreditation is another one that I think is going to be completely renegotiated. And, you know, one of the themes in this course that I was teaching resonated with all the students was that the accreditation system was not really working that well for real quality control. It was not protecting students, you know, from bad actors.

It was not ensuring that innovations get funded. Like that, that to me feels like an area where. I would welcome the reform, and I think many others would, too, for something that is more outcomes driven, that is more oriented towards real impact on students and their lives, rather than the input driven accreditation system that we've got.

So, that to me is just, it's just one area of opening where you could imagine innovators developing new accreditation models that might actually have a chance of breaking into the cartel. Mm

[00:12:07] Ben Kornell: hmm. Mm hmm. In terms of, um, the net funding for education overall, it does seem like the federal government is stepping back and trying to cut costs where you're like, if, if that funding recedes, what are the buckets that will need to step up? Is it philanthropy? Is it, um, state governments? Is it, um, companies or institutions?

Like who do you think fills the void

[00:12:32] Matthew Rascoff: Yeah, it's interesting. I don't know the answer to that. I think it might be different for different institutions because there's so much diversity in the American higher education system. And you know what I say for Stanford, I would not presume applies to other institutions. In the world of research universities, like, you know, research funding has been the heart.

And so, you know, when people say, okay, let's go back to our core. Well, if research funding is being cut, it's hard to know, you know, where the core is and like, like it's our, and, and how much of it there will be. That to me is, it's, it's one of the big questions. There's another category of quote tuition driven institutions that rely more heavily on students and their fees that has not included research universities like Stanford in the past.

But it might be our future. It might be that we need to figure out how to grow the educational offering and, you know, serve more students possibly, and do so more effectively, and, and that, that might be more of our future, and it might be more, you know, diversifying away from some of the traditional federal funding sources.

I think corporate is another one. You know, this is a model. It's widespread at Stanford. It's not ubiquitous across the country, like the industry affiliates programs, MIT media lab helped pioneer this. You know, memberships for corporates, it has certain advantages. You know, my colleagues at the Stanford accelerator for learning have built an education oriented corporate affiliates program that I think is quite successful at translating research from labs at Stanford into products at Google and other companies that are parts of it.

So that might be a more important source, especially in this education area where we are concerned about the translational system. Maybe that source would actually drive, you know, more throughput in, in, in translation because it's more oriented towards, you know, products and it's more oriented towards like commercialization.

So that would be an interesting one to track also in this world of, you know, K 12 and higher ed and, you know, those, those who are doing educational research, if you were reliant on IES previously, you're looking for new patrons and it might be the listeners of this podcast, the ed tech companies. That might provide some of that patronage in the future.

I don't presume to know how this is going to shake out. I just mean like it, the part, the pieces are moving around right now. So we should think a few steps ahead, not just kind of like you said passively, but try to think like. You know, how are we going to play this chess game a few steps ahead?

[00:14:57] Ben Kornell: Yeah, and maybe, you know, to the degree everything you said resonates with me and almost feels like this might be an acceleration point in Fei Fei Li's book around her, her time developing AI. She really commented on how Google had almost become a research institution on par with the Stanford's of the world.

And this crossover of intellectuals between Stanford and Google and back and forth really, uh, signaled a new way of working and because of the capital needed for things like AI research company, large companies and medium companies are really well positioned to actually drive that work forward to not only generate the insights, but then, um, relevant content.

implementation. I think the concern that everyone has if it goes that way is how much of that is open to the public and shared in the public domain versus You know, corporately protected and, you know, kudos to the Google team. Their transformer, um, paper is so seminal in AI. They could have kept that in house, but they published it.

And so that that's the kind of thing that I think could be our future.

[00:16:09] Matthew Rascoff: Yeah. It might also bring, I think, greater separation between research and teaching. You know, part of the model of doing this in the kind of federally funded approach is that, you know, research agencies like NIH, NSF, IES would fund graduate students to be part of a lab.

And then, you know, some of those graduate students. Would become faculty themselves, but some of them wouldn't. Some of them would go off into industry. Some of them would go to SRI and build Siri. And some of them would go, you know, work for Google. And like many of the people who wrote that paper, you know, had training from research universities.

So it's not like we didn't contribute to it. And we just contributed in a different way. And if it is interesting to think like, if that research is really driven by corporate. What, what's the structure that will breed or support or cultivate the next generation of researchers and, and are, is Google really set up to do that training the way that a lab, you know, at Stanford has a kind of pyramidal structure that brings in undergraduates and grad students and postdocs and like, and there's a mechanism for disseminating it, but also kind of training the next generation of it.

And that's not Google's mission, you know, and that, that's, um, that's a concern that I would have if there's really, you know, a research model that moves outside of institutions and into the corporates.

[00:17:26] Ben Kornell: yeah. Well, I think the main takeaway I'm having from this conversation. It's really a jump ball now. And, you know, there's a lot of reasons to be concerned with how the status quo was. So it's an opportunity here for everyone to lean in and figure out what that new future looks like. Speaking of new future, we should talk a little bit about what's going on In a I, um, the main news I'd love to highlight is just the adoption of a I in schools.

Now we're seeing more and more teachers using a I in their lesson plans. Um, new data suggests that up to, um, 80 percent of, uh, Of teachers are using it with their lesson planning. Um, the Ed Week survey said 60 percent regularly use, um, um, AI in their lesson planning. And, you know, the question is how much of the human in the loop is enough and where, where are we actually taking administrative tasks that just are check the box things and letting AI do it?

That sounds great. And where are we actually Supplanting the cognitive lift that teachers need to do. Meanwhile, that report was coupled with a surge in cyber incidents. Um, 82 percent of K 12 schools have recently experienced a cyber incident. My school district, where I've been a school board member, St.

Carlos. We just had a cyber incident through power school.

[00:18:57] Matthew Rascoff: here. Yep.

[00:18:58] Ben Kornell: so, you know, we're seeing this like invasion of AI or, or, or accelerated adoption of AI into the teacher sphere. And at the same time, we're also seeing intruders coming in, you know, from your purview, how do you think a K 12 is going to evolve in these next couple of years with these kinds of dual forces?

[00:19:19] Matthew Rascoff: Okay. With the caveat that I don't currently work in the K 12 system, but I have in the past and I've worked in the assessment part of it in wireless generation, which became Amplify. That to me feels like an area of energy and excitement and innovation. And like, I welcome the disruption to the assessment.

You know, model that we've got, um, especially the multiple choice, you know, driven, you know, part of that, like, I don't know, is anybody protecting like the worksheet as like some sacred, you know, part of the higher education system? I don't think so. I think like, I'm, I'd be okay with letting go of the worksheet model that we had.

Teachers pay teachers may not appreciate that and if people, if magic school replaces it, like I'm willing to live with that, what I would love to see is that, you know, we have an assessment model that allows us to ask better questions that's more open ended, that allows students to respond in a more interactive way that allows us to move away from this kind of like, like a psychometrically obsessed model of, you know, valid and reliable kind of multiple choice and allows us to build like a better formative assessment system that puts the tools Of psychometrics in the hands of teachers so that they can But, you know, open the black box of learning, that's the kind of the classic, you know, power of formative assessment and it doesn't really exist outside of a few small realms of K 12 education.

Early reading has it. Early reading has an incredible system of formative and summative assessment that allows us to deeply understand students who are struggling and students who are on track and to give them the support that they need. Imagine that existed. In science, which doesn't really happen.

Imagine that existed in high school, which generally doesn't have it. And like, to me, like assessment is, is one of the most exciting areas for. For AI innovation. And like, I don't, I don't think the current status quo was working for very many people. So I think magic school had the highest adoption rate of any ed tech product in the history of ed tech.

And that to me is a sign that the teachers were dissatisfied with the options that they had previously.

[00:21:16] Ben Kornell: Yeah. How, um, how is that translating into higher ed? Um, have you seen magic school or magic school equivalents penetrating the higher ed ecosystem?

[00:21:26] Matthew Rascoff: It's interesting. Like it's very uneven. Um, so, um, and I don't think we have a survey like the Adweek survey, I'm not sure that data exists, so it's, it's, it's hard for me to give you kind of a quantitative, more anecdotally, just from my personal experience. Um, the Stanford Graduate School of Business, where I teach, has a new policy on AI, um, that basically prevents instructors from banning it.

For any assignment that students do at home, including take home tests and papers, you as the instructor are not allowed to stop your students from doing that. You may not have a policy that bans it. So we've been banned from banning it, if you think about it that way. This is the first quarter

[00:22:06] Ben Kornell: Talk about two years. What a flip flop. New York Department of Education two or three years ago bans AI and now it's banned from banning. Oh, I love it.

[00:22:15] Matthew Rascoff: Exactly. So, I mean, this, this was the first time I taught under these conditions. Um, I have to tell you, like, I was not using AI extensively in this course. Um, so like, of course, it's, it's something I've been doing with my co instructor for four years now. And Rob Erstein had been doing it for several years before that.

So we, we're not creating worksheets in the class. Many of the assignments are already set. Um, if I were building a new course, I might be using it a little bit more, but it was interesting to see student work that acknowledged the use of AI and to see like the patterns of performance that, that came out of it.

And, and one observation is that, you know, I think there was less variability in the quality, both on the left and the right sides of the curve. I think there was a kind of mean reversion around B plus quality work, you could say, as I told another interviewer. Um, and, you know, the more usage I saw of it, like, there, there were less outliers.

And part of what I'm looking for in student work actually is the outliers. I would prefer that they be outliers on the right side of the curve in terms of, you know, Something really distinctive, but I, I, I don't, I don't love the idea of like compression around, uh, uh, the, the middle of the curve. I'm looking for individual student voice.

I'm not looking for like individual best practices and like, maybe when a teacher is creating a worksheet, they're okay with that. They just want to ask like the standard questions that one would ask on this topic and like, I'm okay with that. But at least in terms of the responses. I, I want to hear something from a student that sounds like them, that sounds like things that they've told me in office hours and in class.

I don't need to hear like the commonly accepted stochastic view of, you know, a consensus of the internet. Like, so like that has me a little bit concerned, I would say in terms of the performance and, and some of the students caught onto this. I told them this, I gave them this feedback and I think they actually may have used AI less as in the second assignment than in the first.

And some of the students, one of them came to me and said, like, I think I'm a better writer than ChatGPT is. And my response to him was like, You actually are a better writer. I've now read your work twice. You are a better writer. So stick with it, you know, because I want to read what you have to say.

[00:24:21] Ben Kornell: yeah. There's there, I think there's a broader question of is AI a raise the floor technology or is it a raise the ceiling technology, and the hope I think, among many researchers is that there's raise the ceiling use cases, but so far we're seeing a lot of raise the floor and on the raise the bar.

raise it higher. We're just seeing that the cut, even the deep research models compared to what human researchers could do from an efficiency to outcome, it might be really good value, but there's pre and post production that really is required to get anything that's raising the ceiling. And, and that may be the limitation of the tech.

[00:25:04] Matthew Rascoff: Yeah. So my, I agree with that. I think that's right. This is sort of the jagged frontier question. You know, it's, you know, the teacher co pilot study that I think you covered here showed that the greatest gains were for, you know, at the floor, not at the ceiling in terms of, you know, tutor performance. It helped those who had the least experience the most.

Um, I also think what we need are AIs that recognize human creativity and productivity more effectively. And I've got this framework that I've been thinking about that is kind of, it's a three part taxonomy for how to think about AI. I think where we are right now is in this one to one model of AI, where it's about like an individual consumer product that's helping me do my work.

It's been conceived as like a personal productivity tool. Okay, so that's like in education, maybe that's tutoring, right? Or personalized learning, something that my parents might buy for me, or that I might buy for myself. Then there's many to one AI, which is about teacher insights. It's about analytics.

It's about dashboards. It's about making those dashboards actionable and usable. It's about red flag systems. So, you know, surfacing data that may be, you know, too plentiful for an individual teacher to process, but that a machine can turn into something usable that they can use to differentiate instruction, the Jeff bus gang study at Harvard business school is a good example of this, where he was mining interactions that students were having with a chat that he built in order to plan his.

Teaching for the next day and he basically used it to surface misconceptions because like he could see the students were struggling with this area or that. So the next day he knew what to support and what to probe and where to push. And then the last model that I'm excited about where I think the creativity might come in where we might actually lift the ceiling is what I've been calling many to many AI and that's about human orchestration.

It's about putting great minds together. In new configurations and allowing people to collaborate at greater scale, possibly worldwide with collaborators that they've never worked with before. Um, you can think of this as like the air traffic controller AI that allows people to work together more effectively.

And Terrence Tao, the mathematician at UCLA has been writing about how he thinks AI is going to fundamentally transform the way mathematics research happens. So you couldn't imagine a higher ceiling, right? It's pure math. research. And he thinks the AI's role is going to be to facilitate co authorship and allow mathematicians to work together more effectively in collaborations, which has not really happened in math.

Other scientific fields have like a big science approach. They're often co authored papers with dozens of co authors. Most math papers are individually authored or co authored by two people who know each other well. And what he, he thinks that proof checking in AI is going to allow mathematicians to widen the circle of trust and work with bigger, bigger groups.

And that's basically kind of the history of the advancement of civilization. You could say it's like letting people do bigger and bigger collaborations, like the pyramids, you know, like the Apollo program, millions of people working on it. Imagine something like that, that really puts minds together in a different way.

And that, that to me is when I think like the creative opportunities and the ceiling lifting opportunities might emerge.

[00:28:07] Ben Kornell: Yeah, I love the framework and it also, um, you know, I think you call it like a human in the loop framework. The idea of it at humanity in the loop is what you call it. Actually. I love that

[00:28:22] Matthew Rascoff: to many AI is humanity in the loop. I think what we have now is human in the loop. I think, I think the many to one is humans in the loop. And where we should head is humanity in the loop.

[00:28:32] Ben Kornell: I love that. Love that. And, and this is where. The strengths of A. I. It can do what A. I. Does really well, which is organizing, structuring data, facilitating process, but it allows the humans to be the spiky creatures that we are and bring, you know, our, our frontier intellectual, uh, possibility together.

[00:28:55] Matthew Rascoff: Exactly.

[00:28:56] Ben Kornell: Um, for those of you who are listening, um, Matthew, maybe we can get some notes for our show notes for those who want to follow up and read more about that.

[00:29:06] Matthew Rascoff: Absolutely. That was a talk I gave at the United Nations, UNESCO Day of Education, which was an incredibly cool experience. We'll talk about it another time. But I, I sort of laid out this framework and with, you know, the ambassadors from all these countries in front of me, and I'm trying to think about, okay, what does AI mean at the national policy level?

I was trying to give them something usable.

[00:29:26] Ben Kornell: Well, so we've, we've, gosh, we've been around the world with the federal government. We've talked about K 12 higher ed and ai. Um, now to talk a little bit about EdTech with that same theme of Raise the Bar, Duolingo published a pretty epic handbook, gave the kind of secret sauce of Duolingo over the past decade and that, or, uh, past 14 years really.

And the, the five key, uh, takeaways were take the long view. Raise the bar, ship it, show, don't tell and make it fun. And what I loved about these five principles and and those of you who can search for this online, it's really a fun read. Um, it really helps us see how the kind of short term incentives around, um, product engagement and user growth.

Um, Can often distract from a longer view of, uh, really creating a transformational learning and ed tech products do lingo really is the, um, unique or unicorn in our space in that, uh, across the pandemic, across all the highs and lows of the ed tech sector, they've continued to just crush it growing year over year, in part because they've Figured out how to do this zone of proximal development in language learning and what you talked about a little bit around the future of assessment.

I feel like that's in the engine of Duolingo is this really powerful assessment engine. So, you know, as you read that and as you've been following Duolingo, what are some of the insights and takeaways that you've got?

[00:31:01] Matthew Rascoff: Okay, so there's more to the story there that I think is in this document. Luis von Ahn is part of what makes Duolingo so special. He was an advisor to my previous team at Duke University where he was an alum, so I got to know him a little bit. He's a MacArthur genius. He's a computer science professor. I think he's now retired or stepped down from Carnegie Mellon.

Um, he is an unbelievable and unique person. And Duolingo as a company is very much the expression of his quirks. His brilliance, his creativity, I think it is, it's a learning science success story also, you know, you were talking earlier about publishing papers and open research, Duolingo has been incredibly active in like the NLP, you know, machine learning, uh, research space.

They go to the conferences and they publish their research. They explain their algorithms. To others who want to use it in, in a way that feels like an academic lab. And it comes from Luis's academic training and background. Like he is a professor first. Do you interact with him? And like that, that is quite wonderful.

They could have just used the AI for SEO. They could have used it in a dark patterns kind of way. Like they are very sticky, like they're very good at that, but they're also using it ethically and for learning and like, you know, zone of proximal development, they're applying all these ideas. That have been proven in labs, but have rarely scaled in ed tech.

And they're showing that they actually work on the scale of millions. And then they publish the research. So they are, I think, a success story in terms of like translation and translation of research, like we were talking about before into products that are joyous to use, that are fun, that are pro learning, that also make money.

And what I hope out of this is like a new PayPal mafia. I hope it's not, I think maybe that's why they're publishing this guide. Like it shouldn't be a one of a kind. It should be that there's many companies like this, but it's hard to name others that have a culture like that. It's hard to name others that have a growth story that have a pro learning growth story, you know, where they didn't have to make compromises on their mission, on their learning outcomes in order to be a successful company and like.

That is, it's, it's wonderful. And I hope, you know, there are alums go on to found many more duolingo type companies. Um,

[00:33:21] Ben Kornell: Um, part of what you said made me feel like this is not replicable because you've got to have this unique MacArthur genius, Luis running things. And, um, you know, also if you follow their story for many years, they had no revenue. And they were just free, free, free, free, free, free, free.

And this was the, an era of VC funding where, you know, they were able to eke it out with VC capital to do user acquisition. So some of that feels like, Oh, that's not, uh, like that's not a well worn path. That's going to be very hard to replicate, but then there's elements of this. Of this playbook and some of the strategies they've using, they've been using that seems super replicable.

So how much of it is this one off unique thing versus this is actually charting a path for us.

[00:34:12] Matthew Rascoff: I mean, the science is all open and, you know, they talk about like, like their marketing strategies are pretty open to their applying behavioral economics. That's basically what they're doing. They've read behavioral economics and they've put it to work to make this really sticky product. So I think it is replicable.

Luis is a unique person. And like the, the company culture is obviously an expression of who he is. But I think that idea of like, like a science driven ed tech model. That does not say that this is other, that does not dismiss it as like useless stuff, as I often heard when I worked in the ed tech sector, like ed schools were like, you know, totally irrelevant.

They don't understand us. Like they don't understand commercialization. Like, no, like there was a. There was a bigger idea there that was about building a better connection, like reading the science, putting it to work, and then contributing back to it. They now have more data than anybody else in language learning.

So they, they're like the world's biggest laboratory for doing learning science. Like, it's not that just that they applied it, they can also now advance it. Isn't that amazing? And so, like, that, that is a repeatable process to me. It's, it's, it's not about the marketing side. It's really more about, like, how you build products in edtech.

And I think it's, you know, my hope is that it can come out of labs. It doesn't have to be faculty who quit their jobs. But it could be, like, the graduate students that I was talking about before, who really understand the science and turn it into something amazing. TeachFX is a good example of this. You know, Jamie, I think he's been on the show, was, you know, is a graduate student here.

Learned from Rachel Lotan, you know, a faculty member here about teacher and student voice about how important that was about wait time. And he built a product that basically represented this research and technology and made it available. And it's now itself, like Duolingo, a research laboratory for Dora Dempsey and other researchers here who are taking that data, putting it to work and then continuing to contribute to improving the product.

Like those, those loops, that's a repeatable strategy that I would love to see more of in edtech.

[00:36:11] Ben Kornell: Yeah. You know, a couple thoughts that are coming to my mind as you say that one is maybe this is part of the, you know, higher ed future model is taking some of that IP and, and putting it into companies. In ed tech that can spin up, you see a lot of this in health sciences and other research fields, but we haven't seen a lot of it in education.

To be honest, the second, the second thing that comes to mind is the purity of Duolingo also has to do with the fact that the buyer and user are the same and it's consumer. And I believe that they didn't have to compromise so many things because ultimately if they were serving the learner buyer, And, you know, if you talk about Jamie, you know, whenever you sit down with Jamie, it's like the product is incredible.

The bureaucracy you have to, the steps you have to go through to sell it is incredibly painful. And so, um, you know, it does, uh, I do think we're entering an, uh, a revitalization of consumer education, uh, one, because, um, frankly, the kind of public dollars and the institutional spend is getting really constrained.

And then two, the available spending from a consumer wallet is going up as people understand, like, My personal and professional journey is much more in my hands rather than trusting my employers or, you know, passively trusting education systems. So, um, you

[00:37:41] Matthew Rascoff: higher ed is very consumer driven. It may be ahead of the K 12 curve on this. You know, because, you know, we have a voucher system in higher ed called the Pell Grant and you can take it anywhere that that's why we have thousands of institutions of higher education in the US, you know, they've, they've proliferated to serve that kind of consumer model.

Um, one of the big topics in the class, you know, that I was teaching was, is enterprise going to be. The key, you know, and, and most, you know, alternative providers, you know, some of the students were observing have ended up with a business model. That's really around enterprise. That's around chief learning officers.

It's around, you know, learning and development. Um, and there's, there's, you know, the traditional higher ed still has this orientation towards. But, you know, the alternative models like Coursera, like Uplimit, you know, these are, these are the examples that our students are pointing to. And there's only a handful of consumer oriented kind of individual purchasing higher ed alternative startup success stories.

Um, campus. edu is one that the students pointed to as an example. Um, so that's, that's like a startup online community college with an incredible, you know, entrepreneur Tadeo Urinde. Uh, he came to

[00:38:52] Ben Kornell: Shaq is an investor. Yep.

[00:38:54] Matthew Rascoff: Yeah. So they're building a new consumer brand, you know, at campus. edu. But that, that's, that's pretty unusual.

It's much more about enterprise and enterprise has the same principal agent problem that school districts do. You know, it's like where the user and the buyer are not necessarily the same. And so like the quality of most enterprise software is not that great as a result. And that is something to worry about.

Like, might we replicate some of those same issues that you see and not have that the same creative energy that you get from the Duolingo, you know,

[00:39:25] Ben Kornell: Yeah. It's probably, um, it's probably not a binary. You know, if you look at Coursera, for example, it started as consumer, but the vast majority of its revenue now is coming from enterprise. Even, you know, look at springboard, like the remaining successful bootcamp out there started consumer, then went with universities now going enterprise.

There's a way in which You know, Duolingo even has corporate clients, so I think you're onto something around like where the scaling factor is. But this, um, principal agent problem, I think It was beautifully solved in Duolingo because it wasn't as big of an issue. And I think that's something that we continue to wrestle with in ed

[00:40:07] Matthew Rascoff: And you can't avoid the inequities that flow from that. Like, if we have a willingness to pay driven model of education technology, we end up with an equilibrium where the rich get educationally richer. And like, that does not feel like a good outcome for a democratic society.

[00:40:24] Ben Kornell: Yeah. Although this is where the vouchers that you mentioned with, you know, Pell grant being a voucher, we're also seeing, um, education savings accounts and other kind of put the power in the hands of the consumer. motions in states across the country. Um, well, we're gonna wrap up with just a little bit of our funding and M.

And a Speaking of government technology, we're seeing a big surge in new companies that are doing public data. A. I. Insights. So Starbridge and Nation Graph just came out of stealth with seed funds. Um, Dan Carroll's backing NationGraph, our good friend Dan, and Starbridge, uh, Owl Ventures is backing them.

Basically what they're doing is they're taking all of the public data from school board meetings and from county office of education and, you know, education commissions and leveraging AI to hoover in all that data and then basically help you understand which contracts are expiring, what are people looking for, what's the kind of RFP landscape, um, coming to be.

And I think we're About to enter an age of precision sales where instead of kind of spray and pray emails and volume based sales, which just really have stopped working, we're going to find way, way more, um, precise, uh, customer targeting that is going to ultimately unlock. More powerful partnerships. I'm actually as somebody who sits on the school board side of things I'm quite excited because our procurement process is totally broken and But our ability to do a nationwide search of every tool or product we should be looking at is also quite limited So these services Starbridge and NationGraph are really exciting to see.

[00:42:12] Matthew Rascoff: When I was at Wireless Generation, there was one person whose entire job was to track RFPs from big districts and states. Like, so however much that person was paid, you know, and he also subscribed to various services that were not AI driven, obviously they were like, you know, syndication services for vendors.

So I think you're right. Like the procurement process does not work for the sellers or for the buyers it's right for a rethinking. So

[00:42:38] Ben Kornell: Yeah, there's there's a way in which Yeah, you might have like a 10 times reduction in cost as well as a 10 times improvement and outcomes. That's the kind of, you know, the impact factor of of AI in in our space. So very cool to see that. And then the other big news is from Sonoma. They acquired learning materials portfolio from Edita for secondary vocational training.

Sonoma is one of the Most interesting low key companies out there. So they're based in Finland. And they've, they started as a newspaper business, acquiring newspaper from non English speaking countries, primarily like Poland, Finland, you know, Eastern Europe. And they ended up getting into the learning space because they saw language as a barrier as a huge challenge.

And now Sonoma, which is publicly traded, I think might be on the Finnish or Norwegian stock exchange. They are using AI to basically, native language and culturally relevant tools, learning materials and curriculum that are super adaptive to all of these non English language locations. And they're aggregating the long tail.

It's a really exciting, uh, business model. You can check it out in our show notes, but I do think there's a way in which, you know, Stanford's always been a global. Uh, ecosystem that there's never been a opportunity larger than today to engage a global audience for you all

[00:44:11] Matthew Rascoff: totally. Um, Jeff Magincalda, former CEO of Coursera came to speak to the students and he said the primary use case for AI, I don't know if he told your audience this, at Coursera was translation. They were able to drop the cost of translating a course, you know, from thousands to hundreds. And so, as a result, they translate many more courses into many more languages.

And like, it's not the sexiest generative AI use case you can imagine. It's not personalized learning. It's not the robot tutor in the sky. It's just making Coursera available. In Romanian, which it's not been before. And I have a Romanian student, you know, in the class. And we were talking about like, that's, that's going to be very meaningful for educational opportunity and mobility and skills in Romania, where not everybody could understand this stuff in English.

So like that, it sounds pretty straightforward, like, but it's actually, I think, going to be very impactful.

[00:45:03] Ben Kornell: Yeah, I mean, in in that use case, and in many others were seeing the. Benefits of AI accruing disproportionately to those who are in developing countries where access to high quality in person education in your own native language is scarce. And so the abundance of this, even if we're talking about raising the quality bar, it is, that's a meaningful shift.

Whereas in the U S because we're accustomed to a high quality learning bar from in person, um, English language, Learning that that may it may not be there yet.

[00:45:40] Matthew Rascoff: And we need a global disruptive framework, because if, you know, disruptions are oriented towards underserved learners, but we're only paying attention to underserved learners in our own country, and if translation really turns out to be as important as it is, and, you know, obviously Duolingo is capitalizing on this also, like this is their whole business, it's like teaching people to learn languages, right, they think languages are still going to be important, like humans are still going to need to be able to process different languages, like we're going to need to have a more global orientation.

For how we think about, you know, ed tech products and like these, they've mostly been conceived as national, but maybe that's just been an artifact of like national funding models. Like, you know, IES, like these agencies that have put the money into it and national policies that have regulated it. Open AI does not respect national policies the same way that like, you know, ed tech vendors do.

So if they become a bigger factor, we might see, we might see more global educational arbitrage opportunities. beyond YouTube, beyond the handful of brands that have really become transnational brands. Like, that would be pretty cool. That would be pretty exciting to

[00:46:40] Ben Kornell: Yeah. And then to articulate your humanity in the loop, uh, you know, argument that ability to translate and connect people across different cultural context also allows us to do transnational intellectual thought work that we've never been able to do

[00:46:56] Matthew Rascoff: That, that was the closing of my speech. That was my effort to say, like, this actually could contribute to harmony and peace in society. It doesn't have to be set up as an arms race. Like, this could actually help people work together, learn together, like, become more cohesive. So, who knows?

[00:47:11] Ben Kornell: on that note, man, I would, you know, the, you're reading the headlines and it can sometimes feel like body blows and body shots, but here, just being able to process all of this with you, uh, has been so inspiring. Thank you so much. Matthew Raskoff is Vice Provost for Digital Education at Stanford. Week in EdTech will be back next week and Alex will be joining us.

He'll be back from his paternity leave, so we're excited for that. all listeners for sticking through this paternity break with us and my cast of co host characters. If it happens in EdTech, you'll hear about it here on EdTech Insiders. I hope to see you all soon.

Thanks for listening to this episode of EdTech Insiders. If you like the podcast, remember to rate it and share it with others in the EdTech community. For those who want even more EdTech Insider, subscribe to the free EdTech Insiders newsletter on Substack.