ASU+GSV 2026 Summit | Tuesday, April 14, 2026, 8:55 am-9:25 am | StarTrack
Speakers
- Andrew Grauer, Quillbot
- Connor Zwick, Speak
- Victor Riparbelli, Synthesia
- Claire Zau, GSV Ventures
Key Takeaways
- This StarTrack panel featured founders of three breakout AI companies -- Connor Zwick (Speak), Andrew Grauer (Quillbot), and Victor Riparbelli (Synthesia) -- moderated by Claire Zau (GSV Ventures), discussing why they chose to build in education despite its historically challenging market dynamics.
- A key debate centered on shipping speed versus product quality: Quillbot reported going from one new product every 3-6 months to nearly one per day, while Riparbelli cautioned that rapid AI-assisted shipping risks creating "40% slop" products and stressed the importance of taste and judgment.
- Riparbelli also delivered a sharp critique of AI industry posturing, arguing "the best strategy in AI for the last three years has been to just lie" through fake demos and influencer promotion.
- On competition with frontier labs, the panelists agreed that focus and deep problem expertise matter more than moats, with Zwick noting "we're not even close to building a truly superhuman tutor." The panel also revealed a striking geographic divide in AI sentiment: marketing AI features actively hurts performance in the West while driving excitement in East Asia.
Notable Quotes
"The best strategy in AI for the last three years has been to just lie. Just create a product demo video of a product that doesn't actually work."
— Victor Riparbelli
"I have a 21-year-old intern that is in our Tokyo office who is not technical and he's shipping faster than most of our engineers."
— Connor Zwick
"Instead of thinking about in really fast iterative cycles of a week, a month, a quarter, how do I zoom out and think about how can I go deep? How can I take the time to really process the information and act with creative confidence?"
— Andrew Grauer
"The more you ship, the more taste and judgment becomes incredibly important... I'm very cautious about not ending up with a product that's gonna end up being like 40% slop in 12 months."
— Victor Riparbelli
"AI will be the most disruptive technology humans ever create... personalized one-on-one learning is the gold standard. It just wasn't scalable before AI."
— Connor Zwick
Full Transcript
We are missing Victor, but I think he will come shortly, but we wanted to go ahead and start the conversation. This is the first conversation of Star Trek, which, as you all know, is a track tailored to investors, other founders, and I am so honored here to have Conor Zwick from Speak, Andrew Grower, and Victor Ripperbelli, the CEO of Synthesia. Instead of spending too much time doing intros, I imagine all of you are quite familiar with their companies. They are very well known in the education space.
In many ways, they are some of the biggest breakout companies of this AI era. I actually want to flip it and ask maybe a hard question, which a lot of people think of ed tech as a very tough sector, where you have slow cycles, pretty price-sensitive buyers, you have a few breakouts, such as yourselves, but I'm curious for you all, as you think about this AI moment, why did you decide to build in education when notoriously people have had so much hesitation around it? I think if you think about the last, I don't know, 20 years, ed tech has been fairly disappointing for many investors and entrepreneurs, because unlike many industries that have been completely transformed in terms of the workflows looking totally different, people are kind of learning in the exact same way they learned 50 years ago, and nothing has really changed. I think that will all be different with AI, though, and that's why we started Speak, I'm sure.
It's similar for you guys, but I think AI will be the most disruptive technology humans ever create. It'll be the most valuable technology, and I think education will easily be in the top five areas or industries where it will completely change the way that we're learning, and I think that's going to be a really big deal for the world. It'll be like rocket fuel for humanity and what's capable. That's why we started Speak.
It all comes down to a really simple concept, which is we know personalized one-on-one learning is the gold standard. It just wasn't scalable before AI, and I think in another five years, it's pretty obvious to everyone in this room that that's probably going to be, hopefully, the way that everyone is learning, and I think that's really cool. Andrew, you've been in education for a long time, and you've been able to adapt and evolve with it. I'm curious, as a founder who's been in it, how you felt about the space, how you thought about pivoting, evolving within the ecosystem.
Yeah, I guess trying to synthesize 20 years is an interesting question, but even when we started 20 years ago, I didn't necessarily think of it as a market of education or ed tech. I think the best way to be building is to really understand in the world what is a problem that is actually meaningful. What is a problem that I actually understand, that people really feel pain about, preferably more frequently in their life, and then I think if that exists, if I believe that, then it's a journey to figure out how do I go solve for that, and I think if you have that determination to go at a particular direction, first started from, hey, people in the world, how can I go solve for you? Then I can be flexible to adapt, and that's painful because my first solution is probably really wrong, and this time, I think my maybe first, second, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth solution might be wrong, but that's okay.
I think if you fall in love with creating value for people and problem solving for people, that's how I think about the opportunity, and I think generally at the highest level, if I can help people, if we can help people, if you can help people learn deeply, create more, do more, better, faster, more affordably, that's an incredible use of my time or your time. Yeah. On the note of, I think, evolving, you know, Victor, when I think about Synthesia, I actually remember seeing it when it was like series A stage, and I had this aha moment. I was like, what would this unlock for film, media, and that was the immediate use case that drew me to it, but then now you all are making such a big push into workforce learning, into enterprise learning.
Curious how you landed on that and how, you know, did you have, did you go into it knowing you wanted to target that audience? It seemed like so many people in L&D naturally were drawn to Synthesia products, so I'm curious, as a founder, how did you decide to, you know, ultimately double down into the enterprise learning space? Yeah, I think we kind of did probably the opposite of what Andrew did, which you're supposed to not do, which is I fell in love with the technology, and then I tried to figure out, like, how do I think this could be applied, and back in 2017, when AI video was, I mean, it didn't work, there was some early kind of research glimpses of it actually working. The thesis for us was mainly just, like, if you could create video from your laptop without cameras, actors, studios, microphones, and it literally would be as simple as it is to create a document or a slide deck, what would that mean for the world, right?
And obviously, the first thing we went to was, like, Hollywood and creative things, because in my personal life, that's what I'm mostly interested in. We then started trying to build for that audience, but quickly realized that, which I think is actually true for many new transformational technologies, right, that the most obvious users are rarely the right first market to target, because we went to all these people who are making lots of great video, the quality requirements was, like, through the roof, right? So we wanted to change the way that they work, because why should they? They already create lots of, like, really cool films and ads, and, like, why care about this AI thing?
So after kind of doing that for some years, we kind of had this realization that there was a lot of people in the world who were not making any videos at all, a lot of them working in the corporate world, right, but they really wanted to make video, because they had this message they wanted to deliver to someone, and they knew that if they share a document, nobody's going to read it, and if someone does read it, it won't stick, because everybody wants to watch and listen to their content. They don't want to read that much anymore. That's just the way the digital economy is going. There's no doubt about that.
In our private lives, that's how most people consume content, right, but then we go to work, and there's a lot of text materials. That was the kind of initial insight that drove us towards the market of, like, corporate video. Learning and training was a great starting point, because it's internal, and it's lower stakes, and basically what the equation was for us in terms of the value we deliver is The really big unlock was that for these people, when they compared an AI video, which five years ago didn't look very good, they weren't comparing it with a real video, which is what almost everyone would do. All investors didn't get it, because they don't have this problem themselves, right?
What these people were holding that video up to was a 15-page PDF document they need to send to someone in their organization to read and understand how to, you know, operate something, and that doesn't work very well, so the crappy AI video was still better, and then, of course, as the quality has increased, the engagement has gone up, but it was kind of one of those things where we just experimented our way to, like, find, you know, find this kind of niche, and I think now, with all the cool things that's going to happen now, we're just, the world is in an upward trajectory of rate of change, right? Like, things are changing faster and faster and faster and faster and faster, and in my mind, there's just no doubt that training and education is going to be one of the most important things, especially in the enterprises, for the next couple of years, right? Everything we know about how we work is going to change, and I speak to a lot of boardrooms and executives who look at everything that's happening in the world, they look at their teams, and they have this feeling like, I don't think my team are AI native, I don't, I'm not sure they know what to do, I don't know what to do with all these AI things, what can I do, right? Like, you, it's very hard for a lot of these companies to go out and hire, like, 24-year-old vibe coders, because they'll, on one hand, be able to join cool tech companies, on the other hand, they'll have VCs throwing lots of money at them, so that's not a viable strategy.
They have to work with the people that they have, right? They have to upskill them, and so what I've seen since we started working on training, which was really back in, like, early 2021, is that it's gone from being a not very important thing in the boardroom to actually now being something that is, like, a top-three concern in many companies, that they have to do transformation of their workflow. It's no longer, like, a nice-to-have, it's, like, a need-to-have if they want to make sure that businesses thrive in the future. Yeah, well, I mean, with the upcoming labor realignment, there is, there are a lot of questions I think a lot of organizations are thinking about what their future workforce looks like, and I would love to get into that and how you all think about that internally with your companies, but I do want to double down a little bit on this kind of question and tension around, you know, you almost started Synthesia from first principles, and, you know, you talk about this comparison, not necessarily to better video, but to existing ways of working, which were very, very much a field of friction, and so as all of you are building products and shipping at such a rapid pace, how do you know when to just incrementally add AI such that your audience, your users actually, you know, understand the aha moment?
When do you start with first principles and introduce something that is categorically new?
feature and product decisions every single day. So in this AI moment, especially as we have a lot of investors, founders in this audience, what is your mindset heading into making all these decisions? I think the way we've worked has completely changed every three months for the last few years and that's accelerating. So I feel like I don't have a really satisfactory answer here.
And I'm definitely frustrated that I feel like in some ways we could go faster. It's very true that, for me, I think the thing that I'm focused on, even more so than whether or not we're building the right capabilities for the long term or whether or not we're solving the user's problems, which are what I would traditionally say are the two most important things in terms of a product strategy. The thing that I'm the most focused on as a CEO is essentially enabling the fastest people in the organization to continue to accelerate how fast they're shipping. Because if you ship faster, it all takes care of itself.
And so we see at the cutting edge, like the top 20% of our team, they are now shipping easily what the rest of the company could have shipped three months ago. They're shipping that in like two or three weeks. And that is different than three months ago. It's like unheard of compared to three months ago, unimaginable.
And basically there's gonna be this giant rift, I think, of how fast different companies are shipping. And if you get that right, everything else takes care of itself. You mean top 20, is that mostly developers or do you have kind of even product leaders, marketers who are? I have a 21-year-old intern that is in our Tokyo office who is not technical and he's shipping faster than most of our engineers.
Wow, and in what ways? He's able to use the tools to build really valuable products and internal tools and other things for Speak. And he's doing laps around people that are experienced developers that are getting paid way more than he is. We're gonna fix that.
But I think it's just a complete asymmetry has emerged in the last three months alone. Are you all, Andrew and Victor, seeing that in your organizations? I mean, at Quilbot, we've been moving from historically a writing refinement business or innovated, and the business was started actually in a similar way, funny enough, to Synthesia where the founders had just watched YouTube videos about how to build with language models and they've just read the attention is all you need paper from Google and in college, we're just playing around this technology with paraphrasing and had built, it's like, what is this? Like, is anybody gonna be using paraphrasing?
What the heck is this? And that's how it started. And as eventually we grew into grammar checking and AI detection, plagiarism checking, writing refinement, writing assistance related markets, we've been now in the last nine months completely moving from just writing assistance into the creativity market. And when we go into the market that is with large businesses from a Canva to an Adobe and many others, it's a huge amount of product that needs to be built.
And you ask yourself, how can one do that? We were shipping historically between 2021 and 2024, probably a new full product with an incredible hit rate of about one per three to six months. We've shipped in the last nine months, almost one per day, which is just mind blowing. And so I do think there is a dramatic change in how much product we can ship and how well we can optimize those per unit time.
At the same time, when you zoom out, I really think there's a huge amount of stress and anxiety that I feel and that I assume a lot of people feel about how fast things are changing. It feels like on a daily basis, a weekly basis. And I think my mind and people's minds have a very difficult time processing exponential change and then assuming it's going to change even faster and that the ground is moving at any point in time greatly underneath us. And so I think my biggest recommendation to myself and maybe it's helpful for you is to try to invert the problem that oftentimes the question is posed as how are you going to move fast and make great decisions?
What if I just simply inverted that when everybody's thinking about, oh shoot, am I moving fast enough? Am I doing this well enough? Is it going to break? I think it makes it really difficult to be calm and to be creative and to be confident and to go deep.
And so I often think about just simply inverting it and instead of thinking about in really fast iterative cycles of a week, a month, a quarter, how do I zoom out and think about how can I go deep? How can I take the time to really process the information and act with creative confidence? Really easy to say. I think it's really hard to do and I think that's very different than what the vast majority of people are able to do in this time.
Yeah, I mean, I think you're obviously we're seeing like the same thing coach and having a huge impact on the rate of shipping. I do think it comes with an additional problem though, which is that the more you ship, the more taste and judgment becomes incredibly important, right? And I think you look historically kind of at like very big companies with a lot of capital. The easiest thing is in the PowerPoint, right?
To say, oh yeah, we're gonna copy these like free products. We're gonna build free individual teams. They're gonna go and do this thing. OpenAI has maybe just gone through and a kind of a cycle of this, right?
Where it's not just always about building more. It has a tax, right? There's an opportunity cost, of course, that may be less now that you can ship things much faster, but there's also a very much like an attention tax. Ultimately, your company has to not just ship features, but ship the right features.
They have to talk about them. You have to ensure that they get adopted. You have to make sure they actually work. And so what I'm very cautious about is like not ending up with a product that's gonna end up being like 40% slop in 12 months.
And that's gonna break the craftsmanship of what we do. I think there's a lot of companies gonna end up with that. But until you've shipped, until like more than half your product is like code gen with no reviewing, you're probably not gonna notice that problem. So I think there is gonna be at some point, there's gonna be a tax to pay for some of the stuff.
That said, obviously we wanna move like very, very fast. But I think it's very important to also remember the taste and judgment part of this. The other thing I would say a bit to Andrew's point is that I think so hard to find a signal in the noise. Everyone right now is posturing to be extremely AI native.
Everyone is talking about this stuff that we're sitting here talking about on stage because we all want our companies to appear to be incredibly fast growing and AI native. And often that's true. But the issue is that it's a strange incentive, right? Now we have a world in which most tech news live on Twitter and social media, but there's no gatekeepers.
And so the best strategy in AI for the last three years has been to just lie. Just create a product demo video of a product that doesn't actually work. You want it to work like that, maybe it will in a year or two. You pay a thousand influencers to pump it for you.
And all of a sudden you've created this perception that industry X is dead because someone byte-coded something in two weeks. But most people don't ever try the product, right? And so you get this anxiety and I'm sure everyone in this room is feeling it. You guys are feeling it.
We're seeing it on LinkedIn and Twitter all the time. You get this feeling like, shit, I'm the only idiot left in the world who's not shipping five products in a week. So I think finding the signal in that noise is super hard. And I spent a lot of my time figuring out that.
I know a lot of big public CEOs who sit on podcasts and they talk about how OpenCloud is their CEO now. And I know that's not true, but they're doing it because the investors want to hear it and the market wants to hear it and their customers want to hear it because they want to work with those kinds of companies. And I don't think that's going to disappear anytime soon, but I think that's something that I try and remind myself of and my team pretty often. It's a bit the same thing with like you're looking at your competitors.
You should definitely look at your competitors, but you shouldn't like run around like a headless chicken and just do what everyone else around you says they do. Because it's always what they actually do. That is really good advice. It made me think a little bit of just this jagged frontier that we're in with a lot of the demos that get shipped from companies.
And the perception that this is what is feasible and every output is going to be as good as what's shown on the demo. And yeah, a lot of users, they try AI and they're like, this sucks. And so I'm curious, are you all seeing actually maybe even a potential tide shift as there is more kind of angst around AI? Is there a hesitation around using AI language, AI productivity, AI video?
In kind of this moment, as you're thinking about, there is a lot of posturing towards the investors who want to see more AI, but then you also see in general masses some angst around this technology. I'm curious how you all are handling that tension. Are you seeing it at all? I mean, I think in, it sounds like in the geographies that you're in, there's less.
of that, but at least in the U.S.
I think we're seeing some heads up. I mean, there's definitely, I think for context, Speak is a language learning app that is, we started in Asia, in places like Japan and Korea and Taiwan, and it's where we are the most popular. We launched in the United States and Western Europe about a year ago, though, and I think it's very apparent to us that the relationship to AI in the West, and in particular, actually, yeah, all of the West is extremely anxious and very wary, and if you just look at the survey data, but in our marketing, if we include anything around AI in the West, it's like, it does not perform, and in fact, you get a lot of negative sentiment, and it's the complete opposite in East Asia. People are extremely excited about the technology, and it's really night and day.
I don't know what you guys' perception is. I don't know. I mean, I think it's interesting, because I've actually, you know, speaking with some founders, they're even, sometimes they say, like, we don't include AI in our marketing, but we just have it embedded in our product, so just as you think about, you know, scaling to masses and marketing your products in this era, like, how AI-native do you want, AI-native, yes, to investors and people who are running companies, but then to your users, how much AI posturing should there be? From our perspective, I don't think showing that we're building with AI is the right answer.
I think back, it's highly correlated to the conversation we just had, I think, at Qobot, when we're moving to help people write, help people design, help people create, ultimately help people create great things, it's, what's the job to be done? What's the problem we're trying to solve? And in this world where I'm trying to be more productive, you're trying to be more productive, you're expected to be 10x more productive, that's anxiety-inducing. That's, there's digital performance anxiety going on, and I think from our perspective, how do we create an experience that give confidence that you can use the best of today's technology, and however it's iterating and changing, we can bring that to you, but for the job to be done, that you want to get done.
Whether that's historically, it's going to be these one-step tools, and if we're moving to more agentic capabilities that can do multi-step jobs, building towards, theoretically, outputs and outcomes, how can we make sure that you feel confident, comfortable, and it's easy to use to feel like you're going in this direction, I'm going in this direction, as I read social media, and then there's a new open-claw or capability, just yesterday. I think there's a big difference between consumer and B2B in this regard. I think in consumer land, right, I think a lot of people kind of instinctively don't like to watch, listen to AI music, for example, even if they may actually like it, like some people will dislike it if they find out it's AI music. I think you see the same thing with AI videos and AI stuff.
Most people probably wouldn't tell you, like, I love watching it, even though a lot of it goes pretty viral, but in other cases, like ChatGPT, right, I don't think I have a single friend that doesn't have a relationship with ChatGPT in a very deep way, they use it for almost everything. In B2B, I think there's definitely the whole kind of automate everything away, for sure has like a negative connotation for very obvious reasons, but I just think that in the work world, right, there is so much incentive for you to be perceived as AI native and to become an AI native that I don't think it impacts it much, to be honest. Like there's so much fear all the way up the chain to the boardroom. The boardroom, right, they're worried about the company getting, like, torpedoed by some AI native startup.
Down on an IC level, you're worried about losing your job to automation. So you are going to be very forward in terms of being someone who masters AI as opposed to someone who gets replaced by AI. But it is very interesting, right? I think there was a recent study where I think AI was like one of the least liked things, like even like politicians was like more liked than AI, which is impressive.
So I think it's definitely a thing. I just think that, I think especially like in the work world, I think the incentives are so big that I don't think it has a big impact. I do feel about what we've seen is that the last couple of years has been a lot of experimentation. So we've bought like a lot of tools to try out things, which I think is exactly the right way to do it.
But now I think there's a bit more of a push towards like, we don't want like nine different point solutions that does different things. Like who can we work with, like what actually worked of these like things we bought? And how can we consolidate that to a few vendors rather than just keep buying like new AI tools all the time? Yeah.
I'm curious just also, you all have done a phenomenal job playing both offense and defense in this moment. As we talked about, the ways that you've approached building your companies, deploying different featured products and decisions behind those. But then at the same time, I think one of the top questions that we often see, and I'm sure you've gone from VCs, other people, Google, OpenAI, Anthropics of the world are all shipping at such rapid pace. Google, I guess, launching language learning or language tools or OpenAI just dominating both productivity and chat or cloud, et cetera, cloud code, spinning up anything or OpenAI, I guess, shutting down SORA.
But all of these players have kind of their tentacles in all these different spaces. How do you think about playing defense in this moment, especially when there is so much attention around how big these frontier labs are getting? I mean, I would say that this is, something's never changed. This was the exact same concern that investors had 15 years ago, 20 years ago.
It was still Google back then. And I think, for me, I personally, maybe this is a bit controversial, but I don't really believe that there is such a thing as a moat, it's just barriers to entry. And this is why you want to work on a really hard problem, because if it's not a hard problem, then it'll be easier for other people to come in and capture the same market value and eventually make it a commodity. So you have to be working in a space that's really hard.
And I think when you do that, then those hard things are the things that prevent other companies from coming in and copying you and stealing the value. And so we're really focused on, we're not even close to building a truly superhuman tutor that can teach you anything 10x faster than the best human in the world. So until we get to that place, we're focused on playing offense and getting there. And we're a lot less focused on what the big labs will do, because we don't even know if we'll be able to solve the problem.
And we think we have a better shot than the 25th best team at opening eye to do it. For what it's worth, I've tried all the tutors from all the big labs, and I don't think we're anywhere close. Andrew? Yeah, I think if you spend your time in this environment trying to defend what you thought you had, I think that's a really big risk.
And I think that's one of the most challenging things, is to say if you have XYZ resources at a large amount of scale historically, it's really hard to then go back to a beginner's mindset in a seed stage mindset, in the sense that, hey, my life has been this box that I've been playing in, living in, or this set of boxes. That can be really powerful, because I can focus. I can go deep. I can explore and understand all of the different crevices of that box.
But at some point in time, you have to also realize that that's a choice. And in this world where we have to zoom out, so much change has happened in literally just three years. And if the beautiful part about software and the internet is asymmetric upside growth, that I could be doing something, and I could get 100x win, 1,000x win, but I might fail 10 times or 100 times, that doesn't matter. That's really easy to say, that for all the venture players in the room, and in just in life, to be able to take this time to be able to learn and be able to play a new box, start small, when I don't have the backwards compatibility problem, it's really hard to do that.
But actually, sometimes that can be totally the right answer, if it's gonna set yourself up for what is probably the largest technological change, for sure, in the last 20 years for me, has happened in the last three years, and it really only feels like high probability that it's going to continue at much more of an exponential rate of change, not even a linear rate of change. And that's still incredibly difficult for my mind to process, and usually a lot of human's minds to process. Sure. Yeah, I mean, I totally agree with Conor, I think this has always been the case, right?
What if Google copies you? And I think the reason is focus, right? I guess, as you also said, you're not competing with Google, you're competing with product team number 87 in Google, with limited resources and lots of corporate bureaucracy around it. So I wouldn't say that's what I spend most of my time thinking.
I do think, when I think about the impact of that on my business, it's actually more about building in a way in which, as the base models and from symbols gets better, your business evolves with it. Because the bigger risk, I don't think it's like an open-air entropic, like going after MySpace in particular. It's more that they enable someone else to get to where I have spent five years getting to with a big team to get there with ten people in six months, right? And that's where it's really important to always think with first principles and really kill your darlings and not be beholden to.
Because we used to train our own models in a specific way two or three years ago, that may not be the right choice today, right? So I think that's more how we think about it. You want to make sure you're always riding the wave, and that you don't get crushed by it. Yeah, thank you all.
This has been such an interesting, I have so many more questions. But I am so excited to continue to see your companies grow. I've been following them since their earliest stages and seeing the evolution in this AI age. And just truly inspired by the ways that you all have evolved and built your companies in this era.
So join me in thanking Connor, Andrew, and Victor. And thank you. Thank you guys so much. Thank you so much.
This transcript was put together by our friend Philippos Savvides from Arizona State University. The original transcript and additional summit resources are available on GitHub. Licensed under CC BY 4.0.