Dr. Jeremy Weisz 16:19
So what was the original use case you’re thinking for this? I mean, obviously now you serve, you know, healthcare, education, global work, I mean, all these companies, a lot of use cases. But at the time, was there something specific you had in mind how this was going to be used?
Kwin Kramer 16:37
You know, the three things we thought when we started out. And it’s interesting. I think those three things are still very fundamental to what we do, even in this new AI era were healthcare, education and workplace collaboration. There’s a there’s a lot more than that. But those three things are such large, important use cases that touch everybody’s lives that they were always top of mind for us at the very beginning, and they’re still top of mind for us today. So if you there’s a pretty good chance that the most recent telehealth call you’ve done was powered by Daily’s infrastructure and developer tools.
We have lots and lots of customers building really interesting things in the education space. So teaching and tutoring applications or learning a new language or one on one support for students in a classroom, and that we should come back and talk about the potential of of AI agents for that, because I think it’s transformative. And then everything we all do at work, if we use a computer for work, we use a whole variety of tools that let us do real time and asynchronous collaboration, ranging from email, which now seems like the oldest of the tools in our toolkit.
Even though I’m old enough, that email was brand new at one point in my career all the way through AI CoPilots that can help us use a new piece of software, you know, 100 times faster than before. So those are sort of the three things that I think about all the time from a business perspective, because they’re so important. I also am super excited about the new kind of social and gaming possibilities of these, of these generative AI models, because we all spend so much time on our phones and doing interesting things on our phones with AI is is something I’m trying to sort of think about how to help people build in really good ways.
Dr. Jeremy Weisz 18:22
What about the healthcare piece for a second? Obviously, how, you know, the telehealth. So like, I mean, you’re the infrastructure that allows this to happen, right? Someone calls those real time audio and video on there and they’re using it for telehealth. How else are is it used in healthcare?
Kwin Kramer 18:39
So I’ll tell you a little story about a thing that really kind of opened my eyes to what the next generation of, of healthcare and I was going to look like. We delivered some voice, AI powered healthcare tooling to a a medium sized healthcare company in, in, in the US, you know, a few hundred doctors and nurses and patient healthcare providers in a, in a system that has, you know, the typical things that all all healthcare systems today do, they have this, you know, electronic health record system where everything is stored and everybody in that system spends a lot of time typing into that medical record system.
And every hour they’re spending typing up patient notes feels to them like it’s taking time away from actually spending time with patients. But you have to do it. You can’t not, you know, do good documentation and and do good health care. So that tension is something that’s providing a lot of push for people to adopt new AI tools that, you know, kind of take you away from the keyboard, back to the patient.
Dr. Jeremy Weisz 19:42
So 100% I have a surgeon friend and he’s like, I, I think I spend more time. It feels like he spends more time doing paperwork than actually, you know, working on patients.
Kwin Kramer 19:52
Sometimes I hear that over and over from people we work with in healthcare. So almost two years ago now, we deployed some voice powered documentation and support stuff for for this, this system. And they were very early adopter and they were worried about all the things you’re worried about as an early adopter. Is it going to be reliable.
Dr. Jeremy Weisz 20:10
Yeah.
Kwin Kramer 20:10
Yeah. So definitely HIPAA and data data privacy. Reliability. Is the system going to hallucinate the way AI systems do and is that going to be a problem. They were very much on the bleeding edge. So we were focused on solving those problems. How do we get you from a proof of concept that you’re excited about to something that you are very confident you can run safely in production. While we were focused on those blocking and tackling very, very important but sort of blocking and tackling problems.
I was in a meeting with a one of the testers, like a doctor in his 60s, loves being a doctor, loves working with patients. He said to me, you know, I talk to your stuff every day in the car. And I was like, in in the car. Interesting. We kind of built these bridges to your health record system, but, you know, they were for, like, specific things you’re doing in the office. Tell me what you’re doing, he said. Oh, like, it knows what my appointments are today.
It knows who they’re with. I just call it from the car. And I talk about what I’m going to do that day. And it reminds me when I saw the saw that this person last.
It reminds me, you know, that they they saw a specialist and, you know, then then they didn’t have a good experience with that. And I had to sort of help them out with that. That happened six months ago. I probably would remember when I see him, but I sure like thinking about that stuff before I talk to them. And it’s really smart. Like I, I ask it some stuff and like it’ll tell me something and then I’ll say, no, I don’t think that’s right. And they’ll be like, okay, yeah, that’s probably true. And I was like that. That is amazing. So this is like a a coworker.
Dr. Jeremy Weisz 21:43
It’s like a high powered executive assistant for someone that remembers everything you’ve done with the patients.
Kwin Kramer 21:49
That’s exactly right. And he viewed it as a thought partner, and he was very clear eyed about what he could and couldn’t do. Like, he wasn’t, you know, he just thought it was a brand new tool that made his life easier and was fun to use in his domain, where he is very passionate about what he does. And that really was one of the very early things in this sort of voice AI evolution that we’ve tried to help out with over the last couple years. That opened my eyes to what the future’s going to look like.
Dr. Jeremy Weisz 22:23
Yeah, it’s it’s there. The possibilities are endless, especially for these these health tech companies.
Kwin Kramer 22:30
I think it generalizes. I think I’ve seen a sort of series of those kinds of things happen, like these emergent behaviors that people get really interested and excited in. And I think we’re just all going to have coaches, executive assistants, co-pilots, Microsoft calls them, that are just baked into all of our interactions with all of our digital systems. And we’re going to view those as very net positive as we as they make the systems we we interact with smarter, more flexible and easier to use.
Dr. Jeremy Weisz 23:09
Yeah. It reminds me, I wonder I’m going to have to make sure. So I had taken tank of the founder of Jotform on and who knows, they may be using you guys, but there’s probably just like collaborations on that front. I’ll send them this episode. It’ll probably geek out on it as well.
So that’s cool. From the healthcare perspective, now let’s talk about education. How are people using it in in education space?
Kwin Kramer 23:38
Lots of super interesting startups experimenting with giving people a couple of different things that I think are are separate. But but they sort of paint the picture. One is one on one tutoring, one on one teaching. The the conversational AI is the sort of best AI models today are very good at having an open ended, back and forth conversation that’s very high quality and that can pull in information from the internet or from a from a specialized knowledge base. That is a really good set of building blocks for a tutor.
So I’m having trouble with math. I’m in fifth grade, I can talk to this tutor and it doesn’t replace my classroom teacher interaction. It just gives me more interactive support. And if you look at education research, one of the very, very replicable results of many, many studies is that one on one tutoring support is always very impactful across every kind of school, every subject, every age, every cultural background. One on one support is just hugely, hugely impactful. And so I think we’re starting to see like it’s coming into focus, a future where one on one tutoring is available to every single child in the entire world.
And I don’t think you can overstate the positive impact of that. So supporting startups and companies that are building the early experiments and making that capability one on one tutoring for every single child in the world available is just super inspiring to me. The other thing we’re seeing is companies building platforms that help people teach things, teach specific subject matter more flexibly and scalably. So I think this kind of vertical platform where AI helps people who are not technical developers build something that’s very capable.
Again, this generalizes beyond education, but we’re seeing it a lot in language learning, coaching, teaching people some specific professional skill, teaching somebody a vocational training stuff. You can again expand the population of people that can learn something really effectively. When you give people this sort of interactive mode, and thousands and thousands and thousands of people all over the world are subject matter experts, but they’re not technology developers. You give people these who are subject matter experts, a way to create a interactive, super flexible, super smart course in what they know best. That is an expansion of knowledge diffusion across across the world.
Dr. Jeremy Weisz 26:10
How are people hearing about you? I feel like if someone’s building a Health tech or an EdTech company, They’re looking at Daily for their audio video infrastructure. How are the companies finding you?
Kwin Kramer 26:24
Typically, one of the things we do a lot is contribute to open source software and the open source ecosystem and for developers, for for programmers, that’s a that’s a pretty powerful way to both hear about a company and be excited and be be confident that you can use the stuff the company does. So we’re the team behind a library called Pipecat, which is the most widely used framework for building voice agents today. So people use Pipecat for things like a customer support call where you pick up the phone, you call your bank or whatever, and rather than sit on hold for an hour while you wait for a human agent to become available, you immediately talk to a voice agent.
And that voice agent can probably do, you know, today, maybe 60 or 70% of the things that the human agent can do for you, which is already a big win in the future, it’ll be able to do 90% of what that human customer support agent can do for you. Or when you call a restaurant in San Francisco, there’s a pretty good chance you’re talking to one of our agents or one of our customers agents to schedule a reservation, because you didn’t need to take the person who’s running around the restaurant trying to make sure everybody in the restaurant is having a great experience and put them on the phone to take the reservation.
The voice agent can totally take the reservation, or a voice agent that calls before your healthcare appointment and interactively, through a voice conversation, collects all that information that you would otherwise have to like, go online and type into a form for. So we’re seeing this sort of huge, huge uptake of these new voice AI agents. All of a sudden Pipecat is powering a lot of that. Pipecat is completely open source and vendor neutral. You don’t have to use any of the data.
Dr. Jeremy Weisz 28:02
I was going to ask because like someone’s going to be implementing, you know, all these restaurants, whatever system they use, obviously the voice agent has to integrate checkout where where there’s tables available. So someone’s actually it just connects with any system essentially.
Kwin Kramer 28:19
Yeah, yeah, yeah. So if you, you know, if you have your phone numbers through Twilio, you can just hook that voice agent up directly to Twilio. If you have a like a lot of hospitals, if you have a super complicated built over many years digital phone system and a bunch of consultants that run that digital phone system, you can hook the Pipecat agents up to that. You know, that very bespoke custom digital phone system. We thought that this ecosystem.
So we started building these voice agents early, sort of before most other people. And we built a whole lot of tools inside Daily to make these low latency, high quality conversations possible with LLMs with AI.
Once we got to a certain point where we thought, oh, this is incredibly exciting, we felt like the best way to help the whole ecosystem grow up around these new use cases was to open source everything we’d done. So that is the origin story of Pipecat is that it was stuff we built for ourselves. We thought it was useful. We thought from every possible perspective it would be useful to sort of put that out in the world with no restrictions on how people could use it. And because we have great infrastructure and because we know how to work with big enterprise customers to help them get to POC and then to production, we would benefit from that open, that ecosystem growing.
Dr. Jeremy Weisz 29:36
So they use Pipecat and they’re like, this is awesome. And like, we need all this other stuff like video and audio and we should might as well just use Daily because, you know, we they’ve created this whole infrastructure. I want to get to the conversational voice because I know you have a lot of what you see is is coming up.
But I do want to hit on that that piece we just talked about, which is you have a very, I think, great value proposition on the website here. When you look at it, it’s almost seems like a no brainer. We’ll see. Build with 10,000 free minutes each month. So talk about the pricing and and maybe, you know, you’ve just it seems like you’ve just tried to create a frictionless path for people to, to use you.
Kwin Kramer 30:24
Well, we’ve certainly tried. And I will say that every startup founder listening to this episode will laugh. If I were to say we’ve succeeded at that because there’s, you know, you’re you’re never done. You’re never done with removing friction, removing pain points for people, getting up and running with your your product, your you’re never done with trying to think about pricing and packaging. We have tried to be really simple and clean about our pricing for our entire history.
So we charge you per minute that you’re connected to our systems. That tends to work well for the majority of use cases. As you know, there are always customers you would love to be able to serve, where any particular pricing you have is tough. The one that’s always been kind of a nut we had trouble cracking is you have a seat based model. Our our customer has a seat based model. They want to make unlimited usage of our stuff possible for their customers, but we charge per minute.
So we do custom pricing for, you know, customers at scale that the standard pricing doesn’t work for because we never want to lose a customer. But I do think that the sort of lots of minutes free we charge you per minute, which generally scales with the value you’re providing, aligns us with our customers. And of course, in the new voice AI world, if you just want to use Pipecat and you want to run all your own infrastructure, we’re totally supportive of that because we think, you know, growth in the whole voice AI ecosystem is great for everybody.
Dr. Jeremy Weisz 31:53
I’m going to pull up Norby for a second because I thought it was pretty cool. Can you talk about Norby?
Kwin Kramer 31:58
I love this project and this startup. And Adrian, the Founder of Norby. So Norby is a little robot that can have great conversations with you and interact with other, you know, computers that are around. And there’s a couple of sort of really obvious applications for Norby. And I also think it’s one of those see around the corner things where as people start to use things like Norby will like, there will be lots of things that we didn’t even think of today that are, that are, that start that look obvious in retrospect.
So the two things that we see Norby used a lot for today are the kind of teaching and tutoring stuff where you have kids and they’re learning stuff, and having a little robot to interact with is a really great form factor for these educational applications, especially for kids that are doing things like speech therapy. So that’s a whole sort of set of things for Norby. Another thing I think about a lot that I, that I think Norby fits into is eldercare. So there are lots of people who are in the in their, their later years who don’t have as many conversational opportunities in their lives as they would like and giving them both a communications tool that lets them connect to other people remotely really easily, which you can do through things like Norby. Norby starting an audio call with a with a human or, you know, a group audio call and who will just be there and have these conversational interactions with you throughout the day is really great.
And voice turns out to be so much easier for so many people than typing that. These new AI models that can have real conversations with you via voice and get things done when you ask them to do things via voice are just transformative for many, many people’s lives. Norby is a startup out of Australia. They’re doing really high end manufacturing of these, of these little robot devices, and they’re building really cool voice first software for them. And they have really big partners like Dell and Nvidia supporting them. So I’m super excited to see where Norby goes.
Dr. Jeremy Weisz 34:02
I mean, anyone who’s doing hardware and software, I feel sorry for them. I mean, it’s such a tough. I mean.
Kwin Kramer 34:09
Yes, but also admiration.
Dr. Jeremy Weisz 34:10
Is hard enough. Yes. One alone. Alone is hard enough. Let alone both of them combined.
Like, you know, I have the aura ring, and I’m like, they have to create this ring that’s indestructible. When I drop it in, it’s gotta. And then the software’s got to work seamlessly, too. It’s just seems like a difficult path. I guess it builds a moat, right?
Kwin Kramer 34:31
No, you’re totally right, and I. I got to know Adrian, the Founder of Norby, because the software for Norby is built on Pipecat. And when I, I think one of my very first conversations with him was about how I have also done hardware in startups in the past, and I say I will never do it again, but I probably will because, as you said, you can feel equally sorry for a founder who’s tackling really tough problems, but also sort of be like, oh yeah, that’s that’s super impressive. So hard, I admire that.
Dr. Jeremy Weisz 35:02
Talk about the conversational voice piece. Where do you see it going? You know, and I know you’ve kind of, I don’t know, predicted this path of, you know, constant video calls. And people were like, no, I’m just going to use the phone. I don’t see this path. What do you see with conversational voice?
Kwin Kramer 35:20
You know, we talked a little bit about how in 2016, telling people that there was going to be real time video, you know, in their lives all the time was something that people just didn’t quite believe.
Dr. Jeremy Weisz 35:30
I remember when I was doing these interviews, you know, I started practicing like 15 years plus years ago, and I would have to get on with someone, have them install Skype at the time, which now I don’t even think exists anymore because Microsoft killed it, but have them install Skype because there wasn’t, I don’t know what else was doing video calls at that time. I mean, obviously there was no zoom and all these other things, so that obviously caught up. So yeah.
Kwin Kramer 35:57
Nina and I, who founded Daily together, talk fairly often about how Conversational voice. I felt very much in 2023 and 2024. Like making that claim about video felt in 2016 2017. We would say to people, look what you can do with these new LLMs. It’s amazing. And they would say, I don’t know why I would talk to an AI, and I would say, that’s not the right framing. It’s not. Yeah, sure. I don’t know why you’d quote unquote talk to an AI either, but I know why you’d like to cut your wait time down from an hour to literally 1.5 seconds on a customer support line. I know why you’d like to ask questions about Math.
If you’re a high school student and you just didn’t understand what the teacher said. And it’s not making sense from the textbook when you get home after school. I know why you’d rather talk to a voice agent that can fill out the form for you before you get to the doctor’s office, so you don’t have to try to pull it up on your phone and type with your thumbs. All the all the answers to these questions. And those are just the obvious things we’ve already built in 2023. Imagine what those things are going to look like once there are millions of developers building things for their vertically specific passions.
Dr. Jeremy Weisz 37:15
Where do you see it? I mean, you say the obvious use cases, you know, making appointments. And, you know, I think some companies are already starting to use it. I know when I call for like a car service appointment, you know, puts me on with an AI agent first. And also, it’s obviously referencing all the data in my file and all the stuff that’s it’s cumulative, it’s smart and it remembers everything. What other use cases do you see besides the obvious ones like customer support, wait time type of stuff?
Kwin Kramer 37:50
Yeah. So I’ll give you the sort of three phases AI. AI predictions are super dangerous. But here’s my prediction. We’re we’re still in the very early days of the obvious use cases like we haven’t built out even 1% of the customer support, teaching and tutoring, you know, sort of online assistance type use cases.
There’s so much to do there. But that glide path is already sort of up and to the right. And I think more people now see that than don’t. Probably next phase is much more integration into all our communications channels. So I often join a video call with four other people, and there are four I note takers in the call that is, that feels very much like an early kluge, as engineers say, on the way to something much more seamless.
So there shouldn’t be four I note takers. There should be individual co-pilots integrated into our software environment that knows each of us individually and can and can be active on our behalf. And I think I’m starting to see startups build those. As you said, AI that has memory, that has agency, and that joins sort of just all of our communications channels. I think that’s the next thing. And then beyond that, we’re seeing social and gaming and real time video avatars kind of be in the same place.
That voice I was two years ago when we were first starting to build the building blocks for voice AI. I think we’re going to have really compelling real time video experiences, and I don’t think anybody really knows what those are going to look like, because we just haven’t had anything quite like that before. But, you know, despite the problems with social media, I get a lot out of things like TikTok and Instagram. Like I learn new recipes that I cook regularly now for my family.
From those platforms, I learn about things I care about, like progress in AI models. These are all videos on TikTok and Instagram that somebody made. And then they put on the platform, and then they get curated by an algorithm for me. But imagine if that video environment was completely interactive. It would be a new thing. None of us have figured out exactly what that new thing looks like yet, but I’m 100% sure that’s the sort of vector of the future.
Dr. Jeremy Weisz 39:59
Yeah. I’m curious when you think of social or in gaming, what are you thinking of in the gaming world? How does that what does that look like?
Kwin Kramer 40:07
Is there a bunch of friends who are building prototypes for new games with AI? And in fact, there’s a really innovative company called Supercell that just launched an AI incubator for games in San Francisco. If you’re interested in AI and gaming, you should check out the The Supercell AI Games Innovation. Brand new, super interesting stuff happening there run by really, really great team. What I’m seeing is three things in gaming.
One is AI production tools are making it easier for people to build games, so you don’t have to be as experienced a programmer or a visual artist or a game designer, because you have these generative AI models that can help you produce the assets, as the game industry calls them. That’s a democratization like we’ve seen in blogging and in video production, and it’s really exciting to give creative tools to a broader, broader range of people. The second thing is game worlds.
Google released a really cool model just last week that is a dynamic game world, 3D world you can walk around in that’s completely AI generated. There’s not code powering that world in the same way as kind of game engines that we use today. It’s all generated by AI in real time. That’s a research prototype today, but it’s obvious that games are going to look like that. And I remember having this experience playing the latest Zelda games with my, my, my youngster and thinking, I.
Dr. Jeremy Weisz 41:35
Remember the gold thing you actually inputted into the gold cartridge. You actually inputted into Nintendo.
Kwin Kramer 41:41
So me too. And I hadn’t really played Zelda since then, but the new Zelda games, the thing that really separates them from my memory of Zelda is just how open ended the world is, but it’s still not completely open ended. You still hit a wall at the end of the world. It’s still it’s still something somebody came up with as a world that you are walking around in. The new AI worlds are going to be as much more open ended as current Zelda as current Zelda is from that cartridge you and I remember from when we were kids.
So that’s interesting. And then the third thing is these worlds are just going to be populated by AIS you can talk to. So we’re sort of looping back around to voice AI, but rather than have scripted interactions with other non-player characters in games, we’re going to have these unscripted AI generated interactions, and that’s going to make the games feel more open in yet another dimension.
Dr. Jeremy Weisz 42:32
I’m curious what other things you see. Before we hit record, we were talking about building more useful things with, you know, with obviously Daily being kind of the backbone. I had on one of the or the one of the founders of Delphi.ai and I, I just saw it said, you know, the founder I actually interviewed his his video clone before I had him on because I was like, listen, you already created this. You did the work. Forget about you. I don’t need to talk to you.
I’ll interview your video clone. So the podcast episode was probably 25 minutes of me interviewing his video clone and like five minutes of him just coming on and chit chatting about kind of how he created some things, but it was really interesting. So I’m curious and I’ll do a few more of those interviews just to see how it goes with someone’s video clone. What else do you see? What other more useful things do you see people using Daily for?
Kwin Kramer 43:30
So those Delphi video and audio clones are really impressive. And that company is really pushing, pushing into the future of what you can do with making somebody’s knowledge more accessible to more people in the world. The those audio and video clones also run on Daily, which makes me happy.
Dr. Jeremy Weisz 43:46
Boom. There you go.
Kwin Kramer 43:48
I think there’s a couple things that that I’ve trying been trying to figure out what I think about and how to help make happen faster. One is how we write software. I’m spending a lot of time with AI code generation tools now, and so are most of the engineers I know. And we’ve gone very quickly from typing out all code by hand, with maybe a little bit of help from our code editor when we hit tab with auto completion to in free form English language telling an AI, hey, here’s the system I want to build. Let’s break that down step by step.
And then you’re going to write the code. And I’m just going to review the code. That is a massive change. A version of those tools is suited for engineers who’ve been doing this a long time. Another version of those tools is suited for people who think of themselves as totally non-technical, but can now write software that’s really great with the help of AI.
What are the building blocks that are needed by those tools to help more and more people of every kind of engineering background or non-engineering background write more software. And this sort of ties back to thing you and I were talking about before we started and hit record, which is I was a baby programmer when the web was new. And the reason I’m a professional programmer is because I encountered the web when I was in college and thought, man, this is going to change the world. And I, you know, was so new at all the tech stuff that I didn’t know any better. I didn’t know why I thought that it was just super compelling.
And so I wanted to learn how it was built. And I went to grad school for computer science after not having after having been a totally self-taught programmer. And that was an amazing journey for me. But the thing, the thing that got me there was the sense that how we build software, how we distribute software, and what software can do are all going to change because of the web. And that’s now true again with AI. The first time I’ve felt that way in 25 years. I’ve loved being a programmer for my whole career, but that feeling I had in 1996, I have that feeling again in, you know, 2025.
Dr. Jeremy Weisz 45:57
I’m curious, you know, obviously you went on to MIT Media Lab and then but you started at Harvard and you you majored in Near Eastern languages and civilizations, so maybe you had no idea. So I’m curious, when you went in, what did you think? You seemed like a pretty motivated, you know, diligent guy. What did you at that point be like? Okay, this is what I want to do after college. Going in with that. Or maybe you had no clue.
Kwin Kramer 46:26
You know, I thought I would be like, a history teacher. I really love history. And I was super interested in, you know, kind of learning history from, you know, going to college and having access to, you know, all these amazing resources to study history. So I, I really loved that college experience of kind of, you know, having professors who knew more than you know anybody possibly could about, you know, niche subjects and broad swaths of history.
And then I got super interested in the web and taught myself to program sort of on the side and applied to go to the MIT Media Lab to sort of blend like that interest in history with, like how you how you build stuff on the internet. And I just got super lucky to get to go to the MIT Media Lab at all.
Dr. Jeremy Weisz 47:07
I was going to say, like, I’m sure that there weren’t a lot of, you know, Near Eastern language majors to who got into the MIT Media Lab. Or maybe there was, I don’t know.
Kwin Kramer 47:16
You know, I will say that one of the amazing things about the Media Lab was how cross-disciplinary it was. There were people doing music and people doing very early AI speech generation and people doing education. I was in a group that did did education tooling, which was sort of the overlap with like history stuff. Initially for me and people sort of building the first 3D printers and the first little embedded circuit boards that you can put in sort of home devices. I just thought the Media Lab was an incredible, incredible place to be in, like 1996, 1997, 1998, 1999. It felt even then, and I think in retrospect, this has held up. It felt like sort of the Xerox Parc of that time and just being in that cross-disciplinary world where everything feels new and, you know, importantly, there’s plenty of funding to build new stuff was just such an incredible piece of luck.
Dr. Jeremy Weisz 48:10
Yeah, it kind of reminds me of I had Chris Heivly, who started MapQuest, and his major was geography. It’s like someone’s like, what are you going to do with geography? Well, I guess he did figure out something to do with geography. It takes you in the path you don’t really expect, I guess.
Kwin Kramer 48:27
And, you know, I think the web opened up those possibilities in a very particular way. And map MapQuest is a great example of that. And now I does that too. So basically anything you know a lot about are passionate about, Have strong ideas about. You can get these new AI models to shape the technology tools that give more people access to your particular domain of expertise.
And the number of possibilities are endless. Like I had dinner with a couple of startup friends last night in San Francisco, and we we just talked the whole night about how much new stuff there is to build, and we just jumped from thing to thing. And, you know, the things we’ve heard from friends that they’re building and things we would like to build but don’t have time to build. And that was like three hours of conversation. It’s just, man, there’s so much new stuff to build.
Dr. Jeremy Weisz 49:15
It’s overwhelming. I mean, even thinking about Daily, if someone pokes around Daily.co, it’s like. The endless possibilities of what you can do with the infrastructure of audio and video and everything available to you.
Kwin Kramer 49:31
Everything you’re interested in and know a lot about. You could hook up to a voice agent and do something compelling with. I think that’s true today, and it’ll get more true over as we improve these tools. But it’s true today. And I’ll just leave you with one sort of big hot take, which is I am now convinced because of how I use computers and how many of my friends who are on the cutting edge of this stuff use computers.
We are going to completely rewrite all of the software we use, and all the devices we use, such that we take for granted that we mostly talk to them and mostly don’t type. And I get I get the same pushback I got about video and about the web. You know, when those were both new, people were like, well, why would I talk to my computer? And my answer is, talking is so much easier.
Dr. Jeremy Weisz 50:20
Oh my God. It’s ten times faster.
Kwin Kramer 50:22
It’s ten times faster. We are we are speaking creatures. And now we have the capability to make software that is voice first in some number of years, five, six, eight, ten years, everybody’s going to look back and be like, oh, of course, of course. We just talked to our computers all the time, and our kids are going to be like, you’re and my kids are now about when I tell my kid like I didn’t have, like I had a tiny little television in our house when I was growing up, like a 12 inch television. We had one.
And he’s like, he can’t even imagine. He can’t even wrap his head around the fact that, like, I didn’t carry an iPad around when I was a kid. We’re going to be exactly the same way about talking to computers. Our kids are going to be like, you had computers, right? And it would be like, yes, and they’ll be like, but you you didn’t talk to them like, no, we couldn’t talk to our computers. They’d be like, that’s so weird.
Dr. Jeremy Weisz 51:10
You know, there was a YouTube channel. I don’t know if you’ve seen it. It’s called Kids React and they put like a VHS tape and box and they’re like, just what do you do with it? And they’re like trying to figure it out. And they explain what it is like, wait, why wouldn’t you just push and go on Netflix? You know, and why do you need to go? You need to go to a store and get a VHS tape. It’s hilarious.
Kwin Kramer 51:33
And, you know, there’s a real sort of aesthetic shift in, you know, company websites and the things kind of some of us are building, like in San Francisco in the middle of this AI revolution, that’s kind of retro tech. Like there’s a there’s a fixed width font and kind of, you know, clean design like aesthetic that, that you see in a lot in a lot of AI companies.
And I think it’s partly that thing where we’re all trying to think about, like, what are the lessons, you know, what are the lessons from the VHS tape to interactive video, like thinking about the technology development cycles of the past is like super, super interesting if you’re trying to design something really new.
I was riding around with a friend last weekend in his electric Volkswagen bus. I don’t know if you’ve seen these, but they’re really, really cool. Totally electric relaunch of the VW bus. Big screen in the front like super super tech forward car. I was like, you know what you need in your electric VW bus that is 2025 model. You need an eight track like this. This car is not complete without an eight track.
Dr. Jeremy Weisz 52:33
I love to talk about team and building the team and the evolution of the team originally. And maybe this coincides with invest investment, right, with Y Combinator. What did the team, the evolution of team look like? And obviously people can check out Daily.co company and learn more about, you know, the company in general. And you’ve had different, you know, investors throughout the years too.
Kwin Kramer 52:59
So on the team side, I think the only general advice you can give in startups is try to hire people who are really intrinsically motivated to do the thing you want them to do. Being excited about the work of the company and the work the individual is going to be doing in that company, makes up for almost every other challenge you’re going to have as you grow a startup and as you work with people for specifically Daily, because we do something that is so deeply technical, we have always biased very strongly towards hiring people who are super, super experienced engineers. Like we optimize the routing of network packets on the internet.
And that’s that’s quite a specialized engineering job. So generally we’ve hired people who, like, have done the kinds of things they’re doing at Daily a couple of times before in their career, and are still really interested and excited about doing it again, you know, in the context that we’re doing it. And that has proven to work out really, really, really well for us. One of the things that’s meant is that we hire we hire remotely. We’re an all remote team, and I miss being in the office with these wonderful people every day. I would love to be an all San Francisco team, but for us, with the expertise we need, the right trade off was build an all remote culture and work really hard on making that all remote culture like super, you know, fun, sticky, productive, all those things.
Dr. Jeremy Weisz 54:30
My last question. I know we have a couple minutes. Is just your favorite apps. It could be computer, phone, maybe those those you can’t live without apps. Or it could be health apps.
It doesn’t matter. You know, I don’t know. I guess on a Daily basis I. I look at I use LastPass a lot and Sweet Process a lot and Textexpander a lot. And you mentioned the voice like I installed Wispr Flow after someone recommended it because I cannot I do not want to type anymore. So I’m curious, your favorite desktop or phone apps that you like?
Kwin Kramer 55:09
Wispr Flow is fantastic. I really like what that team is doing, and there’s much more space to build in these voice areas. As we were discussing, I’ll I’ll give you a hot take, which is that the apps we value the most are apps we’re going to have a complex relationship with because we spend so much time in them. So Slack is in that category for me. I spend so much time in slack and I could not I could not work in the kind of company culture that we have at Daily without Slack.
I also, it’s been too much time in Slack and would like to find ways to work more efficiently even than I can in Slack. So there’s always this. Like you spend a lot of time in an app, it’s absolutely indispensable to your life. You also feel a little bit scratchy about that. I think that’s an inevitable cycle. I feel that way about about Gmail, which, you know, on I couldn’t live without without Gmail now. I couldn’t live without Google Maps now, but they’re both a little bit constraining, and they don’t work exactly the way I’d like them to. I do think in this next generation of AI generated software, all our apps are going to become much more liquid.
They’re going to be much more flexible. We’re going to figure out the building blocks for software that make tools like Slack and Gmail and Google Maps much more tailored to exactly how I use them. And that’s interesting, partly because it’s interesting to me as an engineer, partly because we don’t know what those building blocks are yet. We don’t know what the core bits of software, Lego toolkits we need as engineers to make software liquid.
Dr. Jeremy Weisz 56:42
Kwin, first of all, thank you. Thanks for sharing your journey, your expertise, and I want to encourage people to check out to learn more. If you know someone building something cool with healthtech, EdTech, it could be anything with audio, video, whatever it is robots, robotics check out. So thank you so much Kwin, and we’ll see you next time.
Kwin Kramer 57:05
Thanks for having me on. I’m a big fan.
Dr. Jeremy Weisz 57:08
I appreciate it.
