Edtech Insiders
Edtech Insiders
From Course Hero to QuillBot: Andrew Grauer on Building the Alphabet of EdTech
Andrew Grauer is the co-founder and CEO of Learneo Inc., a leading productivity and learning platform that includes QuillBot, Course Hero, CliffsNotes, LitCharts, Scribbr, LanguageTool, and Symbolab. He also serves as CEO of QuillBot, an AI-powered writing and communications tool. Andrew launched Course Hero in 2006 while at Cornell University and has since grown Learneo into a global platform recognized as one of Inc.’s Best Led Companies.
💡 5 Things You’ll Learn in This Episode:
- The journey from Course Hero to Learnio’s EdTech portfolio
- How QuillBot is reshaping AI-powered writing
- The ongoing tension between learning, doing, and integrity
- Why “translation” is central to AI across text and media
- Andrew’s vision for the future of multimodal EdTech
✨ Episode Highlights:
[00:02:27] Andrew Grauer shares the origins of Course Hero
[00:07:47] Building Learnio’s family of EdTech brands
[00:08:24] Why QuillBot stood out and was acquired in 2021
[00:14:18] Bridging learning and doing through AI-powered writing
[00:20:31] Lessons from two decades of integrity and plagiarism debates
[00:26:18] The multimodal future of QuillBot beyond text
[00:30:39] Translation as the core of AI’s power
[00:37:43] How AI can help learners move seamlessly from learning to doing
😎 Stay updated with Edtech Insiders!
- Follow our Podcast on:
- Sign up for the Edtech Insiders newsletter.
- Follow Edtech Insiders on LinkedIn!
Innovation in preK to gray learning is powered by exceptional people. For over 15 years, EdTech companies of all sizes and stages have trusted HireEducation to find the talent that drives impact. When specific skills and experiences are mission-critical, HireEducation is a partner that delivers. Offering permanent, fractional, and executive recruitment, HireEducation knows the go-to-market talent you need. Learn more at HireEdu.com.
Every year, K-12 districts and higher ed institutions spend over half a trillion dollars—but most sales teams miss the signals. Starbridge tracks early signs like board minutes, budget drafts, and strategic plans, then helps you turn them into personalized outreach—fast. Win the deal before it hits the RFP stage. That’s how top edtech teams stay ahead.
As a tech-first company, Tuck Advisors has developed a suite of proprietary tools to serve its clients better. Tuck was the first firm in the world to launch a custom GPT around M&A.
If you haven’t already, try our proprietary M&A Analyzer, which assesses fit between your company and a specific buyer.
To explore this free tool and the rest of our technology, visit tuckadvisors.com.
[00:00:00] Andrew Grauer: You feel it from first principles in different use cases. You can see it in different studies. You know, there was that like classics Sugata Mitra study where we put the hole in the wall experiment and the kids were learning from each other by just doing right and teaching. And so I think this is an example that's so much more unstructured, but it's powerful that peer-to-peer learning like that.
And especially if it's in the context of I just have a goal or I have a curiosity, that intrinsic motivation to get something done. I think is incredibly powerful.
[00:00:30] Alex Sarlin: Welcome to EdTech Insiders, the top podcast covering the education technology industry from funding rounds to impact to AI developments across early childhood K 12 higher ed and work. You'll find it all here at EdTech Insiders. Remember to subscribe to the pod, check out our newsletter, and also our event calendar, and to go deeper.
Check out EdTech in. Siders Plus 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
for our conversation today, we have an amazing guest. We are here with Andrew Grauer. He's the co-founder and CEO of Learneo, as well as the CEO of QuillBot. Learneo is a leading productivity and learning platform. Offering a suite of productivity and learning tools to help anyone learn and achieve more effectively and affordably throughout the arc of their career.
It's sort of like the alphabet of EdTech contains lots of different D products. QuillBott is a global AI writing and communications platform, helping people think and create great things. Andrew started learning in 2006, launching the brand. Course Hero as a student at Cornell University and has since expanded the platform of businesses to include QuillBott, course Hero, famous CliffsNotes.
We all grew up with LitCharts, Scribbr LanguageTool and Symbolab, which does really interesting. AI for math. Andrew's leadership has resulted in learning being named one of ink's. 2021 best LED companies and Andrew has been recognized as one of Forbes 30 under 30, and an EY at ERNs and Young Entrepreneur of the Year 2022 finalist.
Andrew holds a BA in Spanish from Cornell University. Welcome Andrew Grauer to EdTech Insiders.
[00:02:25] Andrew Grauer: Thanks Alex for having me. I'm happy to be here.
[00:02:27] Alex Sarlin: I am happy to have you here. We have had a few really great conversations over the last few months. I always get so much every time we talk and just learn so much about the EdTech field.
You have been in it for a while. Let's kick off with your EdTech story. You started Course Hero in 2006 as an undergraduate. What problem were you trying to solve back then and how have you continue to expand your sort of ed tech empire for almost 20 years?
[00:02:54] Andrew Grauer: Yeah, 20 years, Alex. At the same time, I feel like I'm at my best when I'm feeling.
Thinking like it's day one. But yeah, I dunno to your question about what was the problem we were trying to solve right back at the beginning. Maybe I'll, I'll ask you a question here. Were you ever overly stressed from homework, tests, papers in school?
[00:03:14] Alex Sarlin: I was at times, yes. I was a procrastinator. So sometimes I over procrastinated more than once.
Yes, often. Pretty
[00:03:22] Andrew Grauer: often.
[00:03:23] Alex Sarlin: Back in those days.
[00:03:24] Andrew Grauer: Well, me too. And as it turned out, most everyone felt that way, unless you're a special edge case, or maybe you just didn't wanna say that, but that's what it turned out to be. And that problem of helping students study was the luckiest and most important thing that we focused on because it's, you know, helping with that stress, that anxiety.
The pain of studying, you know, especially the inefficient and ineffective parts. Whether it was, Hey, Alex, will you be my study buddy? You know, what does that even mean in trying to figure that out? Or raising my hand in lecture, getting to office hours. The social part was awkward. It could be disingenuous.
Definitely inefficient. And I think just honing in and trying to understand that problem, which. But the luckiest and most important part was, it was a painful, real problem. That was, we're talking 2006 in scaling over the next decade and beyond. It was just an underserved problem, especially with online consumer solutions.
And I think it was important to just obsessively learn about that problem and then obsessively try to solve for that problem and all the way, I'd say. A lot of 2020 hindsight bias here, but across the days, months, years, that's what we did and that's how we grew from. Serving students all the way to educators, all the way to professionals from course hero to learning to QuillBot building from learning all the way to productivity.
[00:05:00] Alex Sarlin: Yeah, I was at Cornell as well and I was a few, I'm a few years older than you, and I remember an experience, very quick aside, but I remember experience. I took a philosophy class, I think in my freshman year. And the professor was great, but he had this incredibly thick Scottish accent, just like thick as as stew.
And I remember trying to take notes in that class and just being like, Lou Ava's philosophy run thing is already complex ideas and abstract ideas. And I just couldn't understand a lot of what he said and feeling. And we starting to head towards the tests and the papers and being like, I really feel lost.
And I remember calling home and being like. I am really nervous about this. I've never had an experience like this in high school where I literally couldn't understand what the professor was saying for weeks on end. And they said, go to office hours. My mom told me, go to office hours, talk to him, tell him where you're at and see what you can work out.
And I went and he helped me out and he helped me sort of make sense of what we were doing and I passed the class, but. I wish there was a course hero back then and as many of your other tools as well, but like a CliffsNotes too, but especially a Course Hero back then because what I think Course Hero did and is one of the early, early big ed tech products is sort of give students a sense of community of studying that you're not alone as a student.
That. Struggling to get through whatever you're trying to get through. Especially university. You have a huge cohort of people who have been through it before, you who are going through it, alongside you who are figuring it out, and it sort of was a social solution to a problem that can feel very isolating.
[00:06:26] Andrew Grauer: I love that. I mean, when I hear a story like that is so powerful and motivating, you know, as a builder, and whenever I experience hearing something like that, it's just incredibly energizing.
[00:06:38] Alex Sarlin: Yeah. I think of eo, like the sort of alphabet of EdTech. It's an umbrella company with multiple products. It's not that EO is one product, EO is the family of products, and you include seven.
Different major tools and I know we're gonna talk a lot about QuillBotts, so I'm excited about that one. But Symbolab is a really interesting tool. CliffNotes is an acquisition, a really interesting, literally like a legendary education brand. When you're putting together this family of tools, how do you think about what connects all these products and how do you think about them being a sort of family versus a set of independent products that are connected just under the learning brand?
[00:07:14] Andrew Grauer: Edits core, we really just tried to understand the users, the customers, and the common denominator is there is a overlapping audience with a lot of these brands. And if I think about today and our focus, we currently largely think about two focus areas. One is studying focused brands and use cases. And then two is led by QuillBot and it's our focus on helping people write, helping people communicate, helping people create.
[00:07:47] Alex Sarlin: That makes sense. And you can see the studying, DNA and a lot of these different assistance from lead charts, which does Avada sort of understanding literature to Cliff, notes about making sense of things, sort of synthesizing information to Symbolab. It has all this deliberate practice where you're studying very particular STEM topics.
These are all really hardcore study tools. And then on the other side. QuillBott, which is all about writing. Tell us about this QuillBott, the QuillBott acquisition. So you acquired QuillBott in 2021, right in the middle of the COVID Pandemic, the big ed tech boom. And what made QuillBott stand out to you, and how have you seen it sort of shaping the future of this sort of second branch of
[00:08:24] Andrew Grauer: Eos direction?
It started with just. Meeting an incredible team. This team of founders and a small but growing team was just at the forefront of transformer based models. And they were starting with what seemed to many at the beginning of like just paraphrasing, what is this paraphrasing use cases, even real, and a professor that they were working with out of University of Illinois, urban Band of Champagne.
Would even go to conferences and other professors would ask, what is this thing you're working on? It doesn't have any real world use cases, but, and then he'd be like, Hey, but this quo, look, there's a million people using it every month. It's, I think there's a use case here, and just these recent grads were.
Teaching themselves on YouTube videos out of Stanford and pushing themselves and building these models themselves and just going so deep, having some imposter syndrome, but like pushing through that and just making incredible strides into paraphrasing models and then eventually into grammar checking models.
These transformer based models after like the early Google paper of attention is all you need. It's just like incredible. What they were doing, how they were doing it, and the progress they were making. They had been starting working on this in 2017 and just made just so much progress between then and 2021 when learning acquired QuillBott and we made a big bet on QuillBott.
And it was part of the possibility of dreaming of what an impact QuillBott could make on the world was looking at a company like Grammarly at the time. Right. And what would it look like to build. An AI forward version of Grammarly, what would an AI forward multilingual version of Cobot Cobot being disrupting a Grammarly be?
[00:10:11] Alex Sarlin: Yeah,
[00:10:11] Andrew Grauer: and that really drove the thesis in a way to kind of ground pragmatically, what could we do in the forefront of AI and writing. We've made huge progress in growing Qubo over 10 x across distribution, traffic sales since the acquisition, and that's been great. And now as we've been building CoolBot further into writing assistance, we're looking forward actually to see a huge transformation in CoolBot.
Over the next six to 12 months and that transformation, it's gonna be looking at adding different multimodal capabilities. Cobot historically has been all about text, and we're also gonna be adding more and more different generative capabilities. To help people communicate and create great things is the faster the better.
We can help people create things that's really gonna drive this transformation, and it's gonna be on the web cobot.com as well as building out the copilot where you can use cobot, as you know, across different apps and extensions probably you have the time that this publishes. We'll have about. 5 million Chrome extension cobot users.
Wow. And it following you through all of your different applications, not just on cobot, but of course all the applications you write and communicate in. And part of this, the grounding for us pragmatically, kind of like Grammarly was historically for us, we think that we can build and be inspired by a company like Canva.
But what would an AI forward version of Canva be? That's another pragmatic way for us to see and visualize. What are we going to be doing over the next 6, 12, 24, 36 months? And the guys, you know, I've been jumping back in here to founder mode again as I've taken on the CEO role at QuillBott and just super energized daily now to just try to make this happen day by day.
[00:12:01] Alex Sarlin: Yeah, two things really jump out as you talk about your excitement about QuillBott or many things jump out. But two things, particularly for me, which is, one is you're mentioning how, you know this was acquisition was in 2021. You mentioned they had been working since 2017 and they were working with the transformer models.
They were working with the exact AI technologies that this is the foundation of Chachi, bt, and of Claude and all of these things. So Gobot was really happening. Exactly. In parallel to that, and putting, as you say, AI first thinking about how do we. Lead the product development with these AI capabilities, and that is something in 2021 that was very rare.
Very few people were leading with AI as the sort of core functionality. That's really interesting and I'd love to hear you talk about that. The other one that's so interesting is, you know, Grammarly, I sometimes debate internally about, is Grammarly an EdTech company, right? Companies like Grammarly or Campa for that matter, companies that are sort of productivity tools, they're efficiency, they're about supporting people, but at the same time.
They're supporting you in doing something that is at hard, a learning activity, which is writing and creating. And I think when I'm hearing you talk about Grammarly in this way, or this whole category in this way, I think it's a really interesting piece of the EdTech ecosystem. These tools that slowed in.
There's a few out there that are very popular and they're really helping people in, as you say, through extensions, through following you into your email accounts, following you into your Google creation suite. The idea of being like, oh, it's education sort of at the point of use, and especially around communications and now creativity, that sort of point of use, application of education where like, I need to write a cover letter, I need to write a memo to my employees right now.
And I'm a little nervous about my communication skills and I'm gonna, in real time be able to use a tool like QuillBott to make the best possible decisions to make something really, really, I'm really proud of. I mean, in some ways that's. Education. That's sort of almost one of the purest use cases of education, especially for adult learners.
I'd love to hear you talk a little bit about that. 'cause it's a part of the ecosystem. We don't talk about that much in this podcast. The Grammarly, the Quill. Botts. The SMOs, the SCRs. There's like a few tools out there that are incredibly, widely used. You mentioned know millions and millions of users for education, but we don't always think of them as EdTech.
I'm not sure why.
[00:14:18] Andrew Grauer: Yeah, it's interesting the bridge of when are you learning versus when are you doing Exactly, yes. And when are you learning versus creating. I love that. And I think it's been super fun to be building a product from its beginning that. Is bridging for a use case that goes between student use cases and professional use cases and learning and doing.
I think a great example of that in grammar checking and spelling checking. And we have so many different people that we serve with QuillBot, where we get these reviews and feedback that they get these recommendations. So change this grammar rule, change the spelling, but it's in the context of actually writing an email.
Right? And so you're literally learning. By doing something productive and that dual purpose, and I dunno, in that case, your stack ranking, getting something done fast. But if you wanted to, you can go see the explanation, right? To improve. It's something really powerful for people to improve their writing, but also even language.
Seeing this bridge, a huge use case is people who are writing in their second or third language. Yes. And it's an even heightened pain point and an opportunity with paraphrasing. To really change the fluency and improve the fluency and understanding how this might be written better. So I love that translation.
[00:15:36] Alex Sarlin: Yes. That learning while doing the l and d world sometimes calls that learning in the flow of work. Right? That idea, I think that is a really rich and exciting space for education generally, for all ages. Education. You, you're trying to accomplish something because you care about it, because it matters to you.
And in doing so, you. Learn something. You learn how to do it in exactly the context you need to do it. And looking at the explanation is a great way to really dive down and be metacognitive about it. But even if you don't look at the explanation, you have still learned something, right? If you are trying to write a certain kind of communication, and then a tool helps you improve it, and you make sense of how it's helping you improve it and what phrases it uses and how it fixes your grammar.
If you're a English language learner, or even if you're. Just wanna improve your grammar. The learning is happening in real time, and that is, and we all know that learning happens pretty much continuously in a work environment. You're constantly doing things that expand your skillset. The idea of being able to use EdTech tools to do that in authentic ways is really exciting.
[00:16:34] Andrew Grauer: Yeah. I love that you feel it from first principles. In different use cases, you can see it in different studies. You know, there was that like classic RA study where we put the hole in the wall experiment and the kids were learning from each other by just doing right and teaching. And so I think this is an example that's so much more unstructured, but it's powerful that peer-to-peer learning like that.
And especially if it's in the context of I just have a goal or I have a curiosity, that intrinsic motivation to get something done. I think is incredibly powerful.
[00:17:01] Alex Sarlin: It's huge. It's also relevant. I mean, we saw Claude just this last week launch a sort of learning mode within Claude Code, and it's the same exact thinking, right?
That while you're coding, you say, Hey, Claude Code, create me an app that does this. And while it's building it, it's actually gonna teach you what is it doing, what are the principles behind it, what are the loops it's making? What are the variables it's creating? And that suddenly becomes a learning moment.
And I think you, you
[00:17:22] Andrew Grauer: know, it's really similar here. I mean, that's an incredible change in AI right now, right? Like the fact that in. The language of English, I can code a prototype zero to one in a short amount of time. A design a functional app zero to one. The increase in accessibility that happens from that step function change in AI capabilities is just wild.
I think that's like a moment that is just so inspiring for like increasing access to creation.
[00:17:53] Alex Sarlin: Couldn't agree more. And then it ties back to the idea of if English is the newest coding language. There's like a famous tweet by one of the open AI founders, right? If English is the new hot coding language, then.
You need to know how to do English or you know, you need to know how to communicate clearly in any language, in fact. And then something like a QuillBot can support you in learning how to communicate very clearly, which can then lead into your ability to create your own applications. It's, it's, I think one of the most interesting things about this AI era is how everything sort of has this Lego stacking type.
Functionality where it's like there's inputs and outputs and the outputs become inputs of something else. So if you're sitting in front of coding lovable or cloud code or one of the big application building, no code products, you still need to know how to communicate clearly. And for that you might.
Really find it very useful to have a QuillBot plugin that can paraphrase your instructions and put 'em into a different way. It's just, it's so interesting how these things interplay. The shared language, the language interface, allows all of these different tools to sort of work together. And that's also true of creativity.
There's gonna feel like a slight left turn here, but I have to ask you about this 'cause I just think it's so interesting and so relevant to today's moment in ai. I just have to ask about it. So. Those of us who remember the sort of course hero, there was one of the things that Course Hero evoked in the society, because Course Hero is a tool for sharing notes across learners, across campuses, in subjects or in people who have taken the same class in different years, is a tool for sharing syllabi.
And it was often at the time subject to accusations of. Aiding integrity issues or cheating. They were like, well, hold on. If a student is getting access to the syllabi and some of the notes from other learners who took this class in the past, isn't that cheating? You also have acquired CliffsNotes, which is probably almost synonymous with this sort of vague claim of, of cheating or cutting corners, and now flash forward to 2025.
We are in the grip of yet another. Widespread societal hand wringing about cheating, integrity, plagiarism. What does it mean when students can use AI tools almost every day? There is a big editorial in in The Atlantic or the New York Times or New York magazine about how professors and teachers are just very confused by this moment because of all of this.
I call it quote unquote cheating. We're getting our head around it. You have been in this debate for almost 20 years in lots of different contexts. What have you learned about the integrity and plagiarism world over these years, and how do you address this tension between education, technology being used to get efficiently to an answer versus actually integrating and learning enough to generate your own answer?
Long question, but you get what I'm saying there.
[00:20:31] Andrew Grauer: It's interesting to see the debate at such a large scale with the likes of chat, GPT and Gemini. My core beliefs center around making information, knowledge, education, accessible, like that's how I just get grounded. How do you make education knowledge accessible, all else equal.
That is a powerful thing for helping people. Learn and do more, better, and more effectively. And I think if we can aggregate, we can organize, we can disseminate the best questions, the best answers, the best explanations, all else equal. That is amazing. And I've learned that the incentives and context matter a lot when somebody is interested in a topic.
Intrinsically, the motivation and actions are very different than if they are just required to do something and need to do it and don't want to do it. Mm-hmm. Incentives, context like that matter a lot. I've learned a lot about the context of talk to different types of students with different backgrounds, high school versus college.
Commuter students versus full-time students. I've learned about adult students versus the classic 18 to 24-year-old student, different socioeconomic backgrounds. This context makes a huge difference, is what I found and learned, and one of the other. Top insights is, I've learned there's a particularly big tension when it comes between open access to information versus closed access.
When you're dealing with certain types of assessments and in certain types of assessments, having that open access becomes a huge tension point.
[00:22:15] Alex Sarlin: Yeah. You were sort of right in the heart of this debate way before generative AI was even on anybody's radar, but I think it's, it's a really, it's still. It amazes me how much ink and energy and fear is generated by this concept of, I don't even know how to put it, of students being able to sort of avoid the obstacle course.
That we consider, you know, education like I believe in productive struggle. I believe in having to make sense of things and make meaning. And I, from an instructional design standpoint, I totally understand why the learning often happens with the wrestling. It's not about just getting to an answer at the same time.
Our education system has continuously changed its role societally over the last 50 years, and I think that's both higher ed and K 12, but especially for higher ed, where Course Hero is really operating. I'm saying was because I'm talking about operating at that time, but it is obviously still a huge ed tech tool.
It just feels like that motivation piece you're mentioning is so important now because we've created a world in which we don't measure the learning that happens in universities. We measure the outcomes, we measure the grades and the test scores, and the certification is aligned to that. We really don't measure the learning as much as I think we think we do.
So I think when tools come along, that slightly alter our understanding of where the learning happens, if the learning is happening, give students a little bit more of a leg up in being able to. Complete assignments without needing to wrestle with their way through them in the way that teachers expected them to be doing.
It just the whole edifice sort of starts to fall down. People get, start to get really confused about what education even is, and it creates a huge amount of tension and stress. And we've seen waves of it, and I think you've been, you've been involved in many of the waves of it. It doesn't seem like you are that involved with the integrity issue right now, which is really interesting.
I mean, would you ever imagine acquiring a company for Learneo that would put you right in the midst of debates about integrity these days? Like something that would help students generate answers for their assign.
[00:24:23] Andrew Grauer: I think we're just so focused on helping people write and communicate and create, and I think like fundamentally when it comes to like education, as you said, like if we could just focus on helping people love learning, that's actually at the most root cause issue that you're talking about, learning versus education and never let your school and get in the way of your learning.
There's so many structures. Associated with education in related parts. And yeah, there's a struggle there to learn and try to improve in such a complex system. But at its core, like the dream is to help people learn and love learning. And if you have that motivation, like that's gotta be the focus and the dream.
'cause ultimately, I think like I love the idea if we can use. Technology. If we can use AI to help me think better, if I can write better, and then when I'm writing I can think better and, and recursively, I can then now actually translate that into creating something, these connection points. It's wild to see happening in production.
And I think understanding where or safety problems happen, these are like incredibly important. You need to engage them and understand them and try to solve for them. And at the same time, like understand principally like how do we build for net? Massively positive outcomes relative to the negatives, build for the advantages, and really then trying to deal with the disadvantages and, and real problems.
I think you gotta engage that struggle. You gotta engage that problem as a, as a necessary thing to do while building something.
[00:25:50] Alex Sarlin: Yeah, that's very well put. You put that a lot better than I, I asked it. I think that makes a lot of sense. If education is about loving learning. So you've mentioned the vision of QuillBott becoming a communications platform and a creativity platform.
Beyond writing and text, what does that mean as, when you say there's a multimodal future for Quill, what are you thinking about? And I'm curious how you see the future of AI in general moving beyond text as its sort of core communications media.
[00:26:18] Andrew Grauer: I mean, already we were talking about. You know, when you're saying like, I'm gonna create a resume, I'm gonna create a cover letter.
You know, whether you're creating a paper, writing an email. I think over time, you know, these are different types of creations. They're text heavy. But for us there's gonna be this evolution of different types of communication, different types of creation, where I'm, I want to use different forms of expression, different forms of media.
I wanna create a business card that's text and visuals, some sort of business proposal, a logo, graphs, a presentation, like a brochure, a a menu, a flyer, a poster, you know, social posts, thumbnails. There's so many different forms of expression and creation, and I think if you think about. This relationship again of like helping me think and helping me create, how can we increase the speed from thought to creation is I think like this incredibly powerful thing that's happening.
I think it's so well manifested by in the last six to 12 months. This just change in AI tools for coding where you just increase access, you decrease time from thought to production. And yes, maybe it's working best now for prototyping. Zero one building versus like dealing with an existing code base at this time.
I think that's just an incredible example to illustrate what's possible in the expansion of what we're doing with QuillBott, you know, from text refinement. Summarization for writing use cases, reading use cases, grammar checking and paraphrasing citations, like these are powerful tools to then get into generative text.
And these use cases, we, we really focused historically on refinement as we really have already expanded now, and you're seeing a huge amount of uptake on usage of. Generative text tools on QuillBot, so I can be brainstorming. I can brainstorm an idea, I can brainstorm a thesis. I could take write bullets and expand from there.
I could come up with an help getting an argument or a counter argument or evidence, and then how to cite that evidence. This is like this first evolution from language models that we build for powering text refinement. Into then generative text. Mm. And then what you'll see here over the next six and 12 months, you can already see it in production, the beginning of the following, which is moving into images, PDFs, voice, even visual tools, voice tools, other generative capabilities, ai, what's possible now?
Is allowing us to do so many different types of tasks, and I'm really excited to see how we sequence this set of steps that we go more into different visual types of communication. Mm-hmm. In more generative types of capabilities so that we can be building an AI forward version of tools that have been like really powerful historically of the likes of an Adobe and a Canva.
[00:29:26] Alex Sarlin: Yeah, it feels like one of the biggest stories right now about AI is that AI is so good at translating between languages, which you've mentioned, but it's also so good at translating and making sense of things between media,
[00:29:39] Andrew Grauer: you know, translation's, like a crazy insight actually that we had. You can actually think of grammar checking, uh, paraphrasing as a translation problem, right?
Translation of the tone, translation of the fluency. These literally are translating. One input to an output while trying to retain the meaning behind it, but making that transformation at its core, obviously that's what a translation notes is, you know, from English to Tagalog or English to Spanish and vice versa.
Like these are common thoughts around like what translation is, but I think it's like an incredibly simple but powerful thought of a summarization of expand this text to be bigger Actually. Do a good summarization of shorten this text and turn this whole article into bullet points. These are translation problems, right?
Try to retain the meaning, but transform the input into an output for a different purpose. I think is a really detailed, but like pretty insightful kind of perspective of what is a transformer based model? What is a translation problem?
[00:30:35] Alex Sarlin: It's the heart of what Transformers do, right? The sequence to sequence translation that Ilia Sr.
And the people who created these methods, that's exactly what they were doing, is saying, oh, well, if you took this whole sequence of words and you had to translate it into another language, you're not going word by word. You're not saying this word goes to that word, because then you have no context.
You're getting the sequence of words, making it into a meaning, and then translating the meaning into another language, and you can do that across media. As well, or to your point, you can do that across tone. You can say, oh, I'm gonna write a paragraph and I want you to tell it back to me as a political speech.
Or I want you to tell it back to me as a pirate, or you wanna tell it back to me. As in a song lyric and all it's doing is saying, I'm gonna make sense of each chunk of this. Translate it into an output with a different. Style or a different language. I think in a lot of ways that is how you just said it.
That's the core of what AI is really doing and it's, it's not the only thing it's doing, but it's one of the main things that it's doing is transforming or translating meaning from one place to another, whether it's text to video, text to a different type of text with a different tone. And Quil Bott, I think, is right at the center of this.
So I'd love to hear you talk about how these different media. Activate different ways of being motivated. I mean, we know that young people spend a lot more time with video, a lot more time with gaming, a lot more time with visuals. You know, we saw the sort of Instagram revolution a number of years ago where suddenly it was like the core unit, YouTube, you know, core units of the internet went from being a blog post to being a video or a TikTok video, or a, you know, an Instagram post.
And that's a real mental change. So. When you're thinking about creating a multimedia platform out of QuillBott, how can you help users tap into the right format for the right task at hand and sort of put the pieces together if everything can be smoothly translated from one format to another, you know, how are you thinking about how the images affect users thinking?
[00:32:27] Andrew Grauer: Obviously you can tell I'm motivated by, uh, the way that generative models can help people think and create things better and faster. How it increases accessibility for creating things, whether it's text, visuals, audio, video designs, even software, and I think even when like GPT one or two or I'd say three was coming out, the breakthrough moment like we have to acknowledge was this chat interface.
And it's like for asking questions, generating text now, generating images, videos, audio. It is like powerful to have this conversation with what. Sometimes feels like a person, you know, and that person you can put into different roles, be my tutor, be my assistant, be my teammate. And it's engaging. And I think like engagement all else equal is, is a huge positive, right?
It's motivating, you know, having a personal tutor. We know all of the research behind that. Like if you can increase accessibility to one-on-one tutoring, like in that case for learning, for thinking. For doing, that's like just super powerful and that's a problem that so many people are working on and I think that is really exciting.
I think for QuillBot, I'm particularly interested in what sort of experiments and innovations can we do with user experience, user interactions. Just interaction patterns and models for like moving more fluidly between a chat model and a specific point solution and vice versa. I think we're gonna be working intensely on these experimentations.
On qui bott over the next six and 12 months is exactly that area.
[00:34:02] Alex Sarlin: So when you're talking about creating these user interfaces that are, allow you to move smoothly between chat and other media, it reminds me of a conversation we had with the Google Notebook people at the end of last year, where they have this model of sort of, there's an input window where you can put either a.
Upload material or put a video, a YouTube video or various things. Then a chat window in the middle where you can give it instructions and iterate, and then an output window where it sort of can come out as any different kind of final product. And I'm curious is, is that the kind of experiment you're thinking about doing when you're translating smoothly between media?
Because whoever sort of gets the right solution for how to do this, it feels like it's gonna have a big step up in how we use AI in the future.
[00:34:42] Andrew Grauer: Yeah, really interesting. Yeah. Context management is a huge component of this. For example, right now, if we do a word level or a sentence level recommendation. We have the context of a sentence, it's a different problem and different opportunity to deal with if we're doing paragraph level context or document level context or even broader than that.
And then when we start to make, especially call it like say long form writing recommendations, we have this interaction pattern as you know, where uh, we can show a, a lot of underlines. With either a grammar change, paraphrasing change, different types of recommendations, and that sort of interaction pattern that we have right now is not optimal for all types of recommendations.
Figuring out another innovative, loved by people interaction pattern as we add on different types of recommendations, whether it's generative recommendations, multimodal recommendations, or in this context also like broader recommendations that have involved more context. That's like. Top of mind. Urgent, important problem for us?
[00:35:52] Alex Sarlin: I don't think it's just you. I feel like new interaction patterns that can support our ability to use AI in all of the ways that it is now capable of being used are a problem for the industry. I feel like everybody has sort of inherited downstream The chat interface that you mentioned from chat GBT, people have started to find their own versions of different AI tool patterns.
But as AI is, gets better and better at working with all of these different media, this is gonna be, I mean, I, I don't know if you've played with like a runway or something. You know, the video tools have their own whole patterns that are based on editing software and or Canva has its own whole patterns for how it generates things inside there.
I mean, I think it's an incredibly exciting moment for all of us who care about education and technology and care about technology because it's a new kind of interaction that changes the way we operate. In general, it's gonna be part of everything we do. I wanted to go back to something you talked about earlier, which is just, I think it's so important and it feels like a good node for us to sort of end the conversation on, which is that QuillBott is growing so fast, it's expanding in all these ways.
It's using millions of users. I think it just says an interesting job of bridging between educational and professional tools. You sort of mentioned the idea of bridging between learning and doing. And I think this is something that as an ed tech ecosystem we could really think about a lot more. The idea of, you know, it's not that an education tool has to happen in an education environment, or it has to be purely for, I went here to learn and that's what I'm here for, to learn.
It's. I want to accomplish something and I can learn on the way to accomplishing that thing. It feels like a very different way to think about education and I think it's something you think a lot about. So I'd love to hear you talk a little bit more about how AI can help users move from learning to doing and how the EdTech world can embrace that liminal space between your learning is happening in non learning contexts.
[00:37:43] Andrew Grauer: Yeah, maybe if I just zoom out to what gets me excited in that direction, broadly for education or learning with what's happening in ai I, or just generally technology. I, I do love when, as you said, there's a goal and even better if I like, am intrinsically interested in that goal for X, Y, Z reason, and I think.
Simulation is powerful. I think AR and VR, as we've discussed before, is really powerful in this context. Whether it's like particularly emphasized in cases where you're trying to learn something that maybe is very expensive to do or dangerous to do. You know, let's say like surgery, dismantling a bomb, dealing with an an electrical system.
Mm-hmm. I think these cases. Relative to the classic approaches of read, listen, watch, take a quiz. Being able to simulate visually seeing something and then like doing it yourself and getting feedback while you're doing it is just an amazing opportunity. Especially when there isn't the opportunity to do that in real life simulation.
Real life simulation would be amazing too. And just like from my own experience, it just feels so much better. And then you also have studies that show the relative efficacy being much better. So I think. For the world. I'm personally very excited about that. And one area within that that I also get excited about just personally is related to language learning.
And I've historically just not had the opportunity to really be an immersed environment for speaking a language. And I find it awkward, nerve wracking, dis, I don't know, disingenuous, but like again, these words that I've described about the origins. Of course hero. I think language learning, where I can chat with an avatar anytime I want with no nerves, no worries about how well I'm doing or my perception about me.
AI plus avatars in a simulated environment gets me really excited about helping people speak a language better, more fluently, and much better than the classic Read watch. Listen, yes. Quiz. Repeat.
[00:40:00] Alex Sarlin: I mean that paradigm, that read, watch list and repeat is really born, and I look, I have worked in that paradigm myself.
I've built software around it, but it's really born out of necessity. It's, I don't think anybody given the opportunity to start. Education from scratch would say, you know, how do you learn something? Well, you should sort of remove yourself from the environment in which you're actually doing it. Go to a closed, safe environment with other people who are learning and then read about it in the abstract, and then practice small elements of it, and hopefully you can synthesize those together someday and go back to do it.
I mean, it makes sense. Almost no educational environments. Frankly, it doesn't make sense in learning, in language learning. It doesn't make sense in medicine. It doesn't make sense in anything physical, but we've done it for many years because it's doable. It's something you can teach any subject in the same building, right?
If a classroom doesn't need to have. Machines in it, or it doesn't need to have a a, a simulation or your home environment. Right, exactly. Or your home environment. So I'm with you. I'm so excited about AI's capability to do that. And I would even say that the Chrome extension you've mentioned, you know, 5 million users of the Quill Chrome extension, that is also learning by doing, that's also learning in the flow, right?
If you're in the middle of. Typing a document because that's your work and you're trying to create something that has a purpose and you're motivated either intrinsically or extrinsically, and you have a chrome extension that says, I can help you do this. I can support you. I can take the part of this that you're struggling with and make it easier for you and show you how to do it so that you can make be smoother next time.
Like that. Just like being immersed when you're doing language and talking to an avatar, talking to a native speaker, that is the equivalent of learning by doing for, you know, professional writing.
[00:41:39] Andrew Grauer: I love it.
[00:41:39] Alex Sarlin: Well, this has been so much fun. I always learn so much from talking to you, and I feel like we've covered a lot of ground here.
Andrew Grauer is the co-founder and CEO of Learneo and CEO of QuillBot Learneo. As you know, it's Course Hero, CliffsNotes, LitCharts, Scribbr LanguageTool and Symbolab, which I, I recommend checking out. Symbolab is really cool. We didn't even talk about it today, but Symbolab is great. This has been so much fun and I appreciate you being here with us on the EdTech Insiders Podcast.
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 Insider's Newsletter on substack.
