This Week in Startups - Rabbit CEO Jesse Lyu on launching the R1, future of AI hardware, and going viral at CES | E1885

Episode Date: January 24, 2024

This Week in Startups is brought to you by… OpenPhone. Create business phone numbers for you and your team that work through an app on your smartphone or desktop. TWiST listeners can get an extra 20% off any plan for your first 6 months at openphone.com/twist Scalable Path. Want to speed up your product development without breaking the bank? Since 2010, Scalable Path has helped over 300 companies hire deeply vetted engineers in their time zone. Visit http://scalablepath.com/twist to get 20% off your first month. Uizard.io will help bring your vision to life in minutes, not days with the power of AI. Visit http://uizard.io/twist to get 25% off Uizard Pro for an entire year. * Today’s show: Rabbit CEO Jesse Lyu joins Jason to discuss the viral launch of Rabbit's R1 AI companion device at CES, covering the ambitious product vision (1:37), unique design (23:30), custom-built OS and LLM (11:30), manufacturing logistics (49:45), and much more! * Timestamps: (00:00) Rabbit’s Jesse Lyu joins Jason (1:37) Introduction and Pricing of Rabbit R1, a Pocket AI Device (7:07) Differences between large language models and large action models (10:03) OpenPhone - Get 20% off your first six months at https://openphone.com/twist (11:30) Data collection process for training the Large Action Model (LAM), Neuro-symbolic algorithm vs. traditional Robotic Process Automation (RPA) (16:08) Authentication and access to Rabbit's Web Portal, privacy considerations, and speed and latency of voice interactions (23:30) The role of Teenage Engineering in design inspiration (30:47) Scalable Path - Get 20% off your first month at http://scalablepath.com/twist (32:07) Potential frustrations and limitations with the Rabbit R1 compared to smartphones (41:30) Uizard - Get 25% off Uizard Pro for an entire year at http://uizard.io/twist (42:38) Evaluating the Accuracy of Voice-to-Text Translations (49:45) Product timeline, experimental features, and getting devices to early adopters (1:12:37) Rabbit’s early development, Jesse's journey, and future prospects of AI * Subscribe to This Week in Startups on Apple: https://rb.gy/v19fcp * LINKS: https://www.youtube.com/watch?v=22wlLy7hKP4 https://teenage.engineering https://www.whathifi.com/news/headphone-30-could-a-computer-in-your-ear-change-headphones-forever https://www.youtube.com/watch?v=7iU6K7NccXk * Thanks to our partners: (10:03) OpenPhone - Get 20% off your first six months at https://openphone.com/twist (30:47) Scalable Path - Get 20% off your first month at http://scalablepath.com/twist (41:30) Uizard - Get 25% off Uizard Pro for an entire year at http://uizard.io/twist * Follow Jesse X:  https://twitter.com/jessechenglyu Check out: https://www.rabbit.tech * Follow Jason: X: ⁠https://twitter.com/jason⁠ Instagram: ⁠https://www.instagram.com/jason⁠ LinkedIn: ⁠https://www.linkedin.com/in/jasoncalacanis * Great 2023 interviews: Steve Huffman, Brian Chesky, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland * Check out Jason’s suite of newsletters: https://substack.com/@calacanis * Follow TWiST: Substack: https://twistartups.substack.com Twitter: https://twitter.com/TWiStartups YouTube: https://www.youtube.com/thisweekin * Subscribe to the Founder University Podcast: https://www.founder.university/podcast

Hardware is never a choice by preference. You don't build a hardware because you want to build a cool hardware. Most of the time, if you do things like that, it will fail miserably. A lot of good cases I learned and I experience is that you have a software that's so good that you want a dedicated hardware to make it better. It's always around the software. It's always around what's inside. And then if you have a software that's very, very edgy, very new, you want to de-risk.

on a hardware, at least for the first generation. So to me, R1 is a result of choose hardware by necessity, not by preference. This week in Startups is brought to you by Open Phone brings your team's business calls, texts, and contacts into one delightful app that works anywhere. Get 20% off your first six months at openphone.com slash twist. Scalable path. Want to speed up your product. development without breaking the bank? Since 2010, Scalable Path has helped over 300 companies

hire deeply vetted engineers in their time zone. Visit scalablepath.com slash twist to get 20% off your first month, and wizard.io will help bring your vision to life in minutes, not days, with the power of AI. Visit wizard.io slash twist to get 25% off Wizard Pro for an entire year. That's U-I-Z-A-R-D-I-O-S-T-T-T-E. All right, everybody. Last week, we discussed the Rabbit R-1, which was, I think,

the most notable new product to come out of CES, the Consumer Electronics Show in Las Vegas. You may have seen the demo in viral on X-S-Twitter. Really cool-looking product, a handheld, a pocket AI companion. It has a wheel on it for Navigation. It's got a camera that flips over, and it's got an LED screen, I guess, and is in this bright, beautiful orange, my favorite color.

And it was the bell of the ball. They did a great demo, and I thought we would bring the founder on today to talk about it. It really feels retro in some ways, but it's got a custom LLM, a large language model, and the $200 price point made everybody lose their minds. They've sold over 60,000 of these units already. Over the last five days, they've been selling about 10,000 units a day. So, anyway, the founders here, his name is Jesse Liu,

and the firm is based in Santa Monica, California. Jesse, how are you doing? Good, how are you, Jason? Nice to meet you as well. Did you expect this kind of reaction for your device? Did you think it would become the most viral product, I guess, along with the C-Thru TV. That was the other product that a lot of people

losing their minds about, which I don't even know what the application of a C-Thruz. through TV is if it's up against the wall, what are we going to see the wall behind it? It makes no sense, really, to me. So what was it like? And did you expect this? No, not at all.

I mean, to be perfectly fair with our audience and to be honest with you, we expect maybe 500 units for the first day and maybe 3,000 units for the early adulter market. That's probably it. But we actually prepared to ramp up the order if necessary. So we kind of have a plan B,

but I guess I'm the most conservative one in my entire team. And I think, you know, our marketing team and design team might have a little bit more confidence than us. But me, I'm super conservative. But one thing that I have to mention is that I love the product. The first prototype was actually a Raspberry Pi with the screen. Because we're, I guess we're like super underdog team, and we actually roll out a web version for a very limited amount of users to test like one feature of the lamb, the large action model, which is play Spotify,

earlier last year. And the result is good. And I love the product because I had the first, you know, prototype in this correct phone factor about, I guess, four months ago. And I had the pre-built handmade prototype around eight months ago. And I've been, you know, play with them and carry with them. I love the product. But on the other hand, I'm a little bit in fear that, you know, maybe we're just a bunch

of geeks. And this is like just our own little thing. Well, making something for yourself that gives you joy, at least you know you have one customer yourself. Right. And then you just have to figure out if there's other customers. So I guess let's start off with this crazy price point. It's 200 bucks for the device.

Yeah. And so I looked at that. And there's no subscription. I mean, you do need to have a data subscription if you want to put an LTE card in it, right? If you want to put 5G in it, that's on you. You can just buy, I guess, a Google Fi card for a data. I don't know what those costs. I mean, probably like 20, 30 bucks a month.

They're pretty cheap. How will you ever make money if this is $200 and the hardware so beautiful. How are you going to make a business out of this? I wonder. Yeah. So I guess that's one of the most asked questions from business perspective. First of all, I can tell you that, Jason, we worked really hard to make the perfect balance between design and bomb. And even though I cannot give you a very exact number of the bomb, because I'm not allowed to share that, but we're making money out of the hardware. But given by what I observed and what I learned in my past careers, the hardware margin

is so low. You know, if you look at a phone, you're looking at negative 25% to 7%, 8%, that's it. Still, most of people, they're trying to make money out of, you know, hardware, obviously subscriptions. The choice of making the same tray in comparison of a e-sim is both reasons. is one, we want to further cut down the budget the bomb cost of the device because it seems require

a more expensive parts on it and same tray is literally just a tray. But more importantly, is that we want to sell to multiple destinations instead of, you know, we have to have a negotiation with a career like in US. We're talking about Verizon or T-Mobile or ATT and stuff like that.

We just move too fast and there's no way for us to sit down, you know, a year to talk with these guys. And now they're reaching out to us, which is great. Fantastic. Yeah. So, so, so, so it's all, it's all strategic decision. But I think, first of all, I want to correct one thing. I actually watched your last episode and talked about us. Yeah. We're not making any large language model. We, we, it's, the large action model is neuro-symbolic. It's not a large language model. Large language model, where we usually talk about

is GPT4 and Bard and, you know, GROC and all that. Those are based on Transformer, and they require a ridiculous amount of GPU on the cloud to be able to train and get things right. I don't think any startup raised $30 million can just out of a sudden make their own large language models. So for the record, we're not doing that. We're working with all the best language models

and small language models, and potentially if there's open-source language models, models that's in the future, we would like to basically set up an EVO, an internal evaluation to keep monitoring the performance from all these major vendors and we can switch on the fly. That's how Rabbit OS works. But we are focusing on large action model, which we know for a fact that the language models, the transformer, were designed to understand language better, but it performs really, really miserably at least for now on finishing the tasks.

And we don't like the way to work with APIs because API, when you're betting, everyone will give you API, which is not the case. It's easier for Open AI to encourage or big companies to encourage everyone to build API for them, but it's really hard for startup to convince, you know, all of a sudden you have 2,000 vendors working on your API in your format. But even if you have all the API,

oftentimes they don't 100% replicate the full feature of the application. You know, say I worked with Open. Uber API before, back in my previous company, Raven Tech, 10 years ago, their API can do three out of 10 things that you can do. And it's really hard to convince them, say, hey, can you do all the things from their app? They have no incentives. So we don't like the API. That's why we were like, okay, hold on.

Let's take a step back. Let's think about a universal solution. Let's create an AI for a universal solution to whatever the application is, is Android app, iOS, Windows app, how can we build a universal solution? Knowing that language model weren't designed to trigger actions. So we actually use Neurosymbolic

to basically start, we actually start working with data labeling companies. We started to collect, based on our own assessment. We actually assembled test groups to have real human interacting with different kinds of software, like Uber, like Spotify, like all the frequent apps. So we started this process about two and a half years ago. we started collecting real human interacting with all kinds of software

and we get all those things in video recording. And then we set up a neural symbolic algorithm, which become today's large action model, that you basically feeding all these clips into the large action model and ask our land to read the clips frame by frame. Are you still using your personal number for business? Well, stop such a common mistake that founders make, but you never have to make that mistake again

because of Open Phone. Open Phone has rethought every single detail of what a modern business phone should look like. Open Phone makes it super easy to get a business phone number not only for you, but for your entire team. And here's the magic. It works through a gorgeous app that works on your phone and your desktop. I can tell you open phone is amazing because all of our sales team and ops teams use it every day. Why? I don't want people using their personal number. Then they leave the company and they're still getting phone calls from our customers, clients, and partners. No, I want all of that to be professional and open phone. Is the number one rated business phone on G2 for customer satisfaction because it's so professional and easy to use?

Here's a feature I love. You can create a shared phone number with multiple employees fielding calls and texts from that number. So we want to reply to our founders, to our partners really quickly, and we don't want to miss a call. We don't want to miss a text. And that's why we use open phone. And it's already affordable starting at just $13 per user per month, but twist listeners

get an extra 20% off for the first six months at openphone.com slash twist. If you got existing numbers and you're paying through the nose for some insane service, you can port those right over to open phone at no extra cost. So here again is the offer.

Go to openphone.com slash twist and get this all organized. Get the 20% off as well. Openphone.com slash T-W-I-S-T. So you would record me using Uber Eats to order my sushi or to order my car. You record a... enough people doing that over time,

then the large action model, the lamb, is understanding the pixels on the page, it knows it's an app, and it knows where I'm clicking in the app. So when you have the rabbit and somebody talks to it and says, hey,

order me an Uber black, and by the time I get home, I would like to have some sushi for a family of five ready to go, give me a range of things, including some rolls and some vegetarian options, It would know how to do that because you've trained it so many times watching hundreds,

I don't know, thousands of interactions in a singular app. How many interactions do you need? Did I describe correctly what you're doing with a large action model? Yeah, yeah. You're described in a correct content. First of all, we don't require from you. We have a test group that we assign the tasks. You know, we actually work with data labeling parties.

So all those groups were purposely connected without violating anyone's privacy. privacy. We're not like setting up something to record on your logo. We never do that. But yeah, the idea is correct. We actually had the entire paper published on the rabbit text slash research with the actual, you know, behind the scene stuff. You can go there and take a look at that paper. But yes, we collect from real human interacting with this apps. And what happens is that it's very ironic because Neurosymbolic runs better on CPU than GPU. So our cloud host is not as crazy. Merely insanely compared to Open AI compared to any large language model,

our cloud deployment is very, very reasonable. We're not talking about hundreds of millions of cash. We're not even talking about tens of millions of dollars. We have a good enough GPU cluster, and we have a good enough cloud computation in CPU. And what happens is that we're not collecting per request. We're just asking people to play around. For example, the way we collect the data.

So the task being deployed, be like, hey, you have 10 minutes on Spotify. Try as much as you want and do as much as you want. I'm not going to tell you how to play this track and click here and do that. You explore free flow for 10 minutes. And the neuro-symbolic algorithm, one of the biggest difference, between the traditional RPA, for instance. If you're familiar with RPA, it's like basically you're recording the screen, of course, but you are then deploying an algorithm, a pre-programmed sequence to navigate your mouse,

your cursor, to the X, Y location based on the absolute coordination. R. R.R.P.A, you're referring to robotic process automation? Exactly. Yeah. So I'm trying to say that. So people understand when you're programming a robot, you can actually take the robot arm. Yes. And move it to pick something up, put it into this box, pick a different object up, put it to box B, it records that, it learns that, and then it can just do that over and over again. You've literally done the task. It's almost like showing a monkey, you know, how to peel a banana,

and then it peels it the way you do it. Monkey C, monkey do, basically. Right. So that's RPA, but Neurosymbolic take one step further because we're not recognizing all those elements by absolute coordination of the entire ratio of the screen. We extract and auto-label some reasoning about elements directly from symbolic method. So that means that it doesn't really matter if a app completely re-ramped the UI. Got it. So when Spotify redos its app and it moves podcasts from tabs and then it puts the tabs at the lower part or puts it in a hamburger drop-down, it still knows that's the word podcast and that's where you find podcasts inside of Spotify. Exactly.

Because the fundamental philosophy logic is that all these modern software are designed for human eyes to proceed information and they have to have a setting somewhere. They have to have some symbolic thing and text somewhere and search bar somewhere. So that's the advantage for us in comparison with, let's say, just build a hardware on top of GDP4. So I first want to clarify we're not working on any LOM. We work with LOM, but we also created LAM, which is this neurosymbolic method. So you have some bank of people in Manila or offshore somewhere with these apps playing with them every day. that's been done. That's all been done.

So you've trained it on. Yeah, I've trained. Let me ask a question there. When you, when I tell Rabbit, hey, order me sushi, family of five, you know, this amount of food, et cetera. Yeah. It knows how to use the Uber Eats app or the DoorDash app. Now the Rabbit device then fires off to your service somewhere in the cloud.

Yes. This request. I have already authenticated in a web interface. Right. hey, I authenticated my Uber Eats and my DoorDash account. It knows the sushi restaurant I like. And then it starts this ordering process. And then I guess comes back to me with like, hey, just want to confirm this is what you want.

And I say yes. And then what does it do? It pops up an emulator of some type in the cloud. And then you have this web emulator that has my login authenticated already. How does that work? So first of all, let's start from beginning, authentication, right? Because if you think about the device, it works quite different than the previous generation. Because it doesn't have any software pre-big thing.

It doesn't have anything pre-installed. It's literally just an AI. And it's your choice, what kind of service you want to enable. And it's your choice, how complicated and how advanced you want this device to be. If you just say, hey, this is a cool-looking iPod, I just want to use this to listen music, then you unlock the music feature, choose whatever vendors. It only will do music.

But tomorrow, if you want to start ordering food, you have to unlock that feature. And the logging process is where you're correct. We have a web portal. That's kind of like our own mini version of iTunes slash ICloud, if you understand that in that sense. That helps with all the authentication settings and feature management.

So you go to the website and you basically choose whatever service you want to unlock. Because again, to large action model, Spotify, YouTube music, Apple music, doesn't make a difference. It is an interface. In fact, Expedia and YouTube music

doesn't even make a difference. They are all interfaces. So we give you the freedom to choose whatever preferred services you want to unlock. You go there and you basically click the Connect with Spotify button. And what happens next is that it will redirects you to Spotify login.

and we don't save your credential. We don't touch that. You go to Spotify, you go to Uber, you go to DoorDash, you log in through them. And then we recognize, oh, this account is being connected with Rabbit OS. And then what happens is that on our cloud, we have a very, very creative structure and innovative structure that we have considered we have a superhost. We have a super host computer that when Jason talks to his Rabbit R1, about ordering a hamburger from DoorDash,

what's going on is that we first see if Jason's logging with DoorDash or logging with Uber Eats, right? And we saw, okay, Jason chose DoorDash, and then on that super host, a Lam is interacting virtually with a DoorDash app or website, an interface. And you don't see all of that because all of a sudden this is done all the ones because it's AI.

And then we will re-render rabbit, themed E.O. On your device to get your result. Got it. So you're not

directly interacting with the host as well. You're just talking to it. The intention goes LOM, I understand,

oh, you want to do this, then LAM do that in the virtual environment and re-render the result to your device. That's how it works.

And so, do you need permission from Spotify to do this? Or you could do this with any app once you've trained it on the

data. I mean, it's nice to get their permission, but it sounds like you could do this without. It is not. It is definitely nice to get these guys permission or I wouldn't say permission. I think we should develop a better business model because this is to me, it's like, okay, you know, remember the early days, Steve Jobs called Sony, be like, hey, from tomorrow, because we have this device from tomorrow, 99 cents per track, you know, I feel like this

kind of like a similar situation because we are not, first of all, we're not creating new users, right? We're not creating spam users. We're not creating prepay users. We're not creating prepay users, you are you, you're Jason, under your authentication to their interface to use their service. Whenever, however the way that you ask, as if you're using it on the phone or on the TV. So we take a good amount of time to studying the terms and agreements and trying to understand this. Seems to us that unless they shut down their interface, which is not going to happen.

because we're not violating anything and we're not even create fake user spam. There's a lot of ways we can do that, right? You know, I'm not sure if you saw the nothing, sunbird, I message bypassed through Android, you know,

there's a lot of weird way that you can set it up, but we're not setting up any of that. You are you. Yeah. And it makes a lot of sense because for them, this is just their user, interfacing with their app through essentially a voice interface with some confirmations on the screen. Explain to the users who are listening, like as a user, I order that hamburger through DoorDash.

It says five guys. Or does it give me like a choice? Say like, hey, you've ordered from five guys and shake shack before. Where would you like to order? Would you like your standard order? Would you like this? Would you like to repeat an order? How at this point in the 1.0 of what you've released?

or the point one of what you released. How does that interaction work? What if the restaurant is taking two hours to deliver or they've stopped delivering? Right, right. Yeah, that's a great question. I think that's something that we're really focusing on because this is a new problem to solve, right? So one part, we know for a fact that we will just directly trigger the service if it's a direct action. You know, if you want to listen, get lucky.

That's it. You know, get lucky. You want to be to be there right away. that's easy. But we also identify there's a couple of scenarios. I guess one thing or reason that I didn't show on my keynote

with the B-Row on the travel, I booked the entire travel plan through Expedia. Some serious internery stuff or some serious stuff relates to, you know, I guess a lot of deposit. We basically want to create a way for you to

check back over and over again. Confirm, confirm. Exactly, yeah. Everyone's meeting about that. I'm Mimi about that myself. So that's why I think we need to separate categories. There are certain categories. You want to get a Uber to your

home, right? And that's it. That's simple enough. We get it. But if it's more complex text in the future, you should be able to see a copy of what exactly going on with the actual paperwork on your rabbit whole web portal. That's how we want to design it. So you always get a

portal to check more legitimate stuff for backup and stuff. And you're no taking, you know, your summary of the meetings, all those stuff will be synced to the white portal. And that's why I call it more like our mini version of ICloud. But on the device, because it's so tiny, I think a lot of people, because I have a ridiculously small hand, I think a lot of people get wrong about the size. So it's half an iPhone.

It's precisely half an iPhone. Oh, it's precisely half an iPhone. You're right. It's same thickness of an iPhone 15. It's a max model. It's a bigger model. But, you know, I'm pretty much. this next to it.

Yeah. Because it's such a small device, we actually had a hard debate whether we should just cancel the screen at all, right? Because that's a decision. Like the iPod shuffle where it's just, you know, like really, really tiny. So it just becomes like a tiny recorder or whatever. You could make it into a watch, you can make it into any number of things. Exactly.

Take me around the form factory. You used a really great design firm to design this for you. Maybe you could talk a little bit about who you used to design it. who you use to design it and the inspiration behind the device, because it looks modern and retro at the same time. Yeah. And it's iconic. Yeah.

I don't think I'm confident to say it's already iconic, but I saw a lot of people are making free works for us, for the case and for everything. I think that's a good start, you know, at least. It's a great story between me and teenage engineering. I respect them. They're my hero company. I started collecting vintage synthesizer about 15 years ago.

So when I saw the OP1, the portable synthesizer they introduced, I made it about it. But it's very, very expensive to me a couple of months to get the findings together in the beginning. But first of all, teenage engineering is not a design form. I have to make clear of that. They are not a firm that actively looking for collaboration and taking design fees and do that. They are a very focused, you know, music, technology, consumer firm that is being a very focused. over, it's been there for almost like, what, 60, almost 20 years, 15, 16, 20 years.

It's such amazing talents of groups of people that I told my team when I was in Raven, I'm like, hey, if these guys gave me a call and asked me to go work for them, I'd go, you know. And then fast forward, when I was working on Raven Hardware Project, which is Raven Asian Raven R back in 2017. I'm like, okay, hold on second. Maybe I can convince them to work with me instead of I work for them

because I want to pick the best and they're in my mind the best. So I literally just call the email, reach out, and three days later, I sit in Stockholm at their office. And then Jasper works in. Jasper is a CEO and co-founder of Teenage. And we both had an app at a pencil. and we started to, you know, start to draw things. And during that process, he asked me about, oh, what's your favorite artist?

What's your favorite car design? You know, and a couple of questions, like wipe chat questions, right? And it's quite surprising to me that we have the exact same taste on almost everything. Like all the questions he asked, he showed me his vinyl collection. I showed him my vinyl collection. It was exactly same band in the same sequence. And he said, oh, I love this and I love that. And we talk about a couple of the things.

So they're like, okay, let's do this. The whole process is like magic. I don't think I ever talked about this on any media or social network. But we actually set it up a secret Instagram account. We have no email communication, no phone calls, no nothing. We just post sketches on the visuals on Instagram. And then we just like, he likes, we leave comments. That's it. That's it.

So we did that. Whoa. We did that. We did that Raven Age project. So no Slack Connect room, no I message group, a secret Instagram account, the teenage group post to it. You post comments, et cetera. Is it a public account or is it a private account? It's a private.

A big comment account. What an interesting way to do sketches. Yeah. I think because it's so long, maybe later later down the road, I might be. be able to share some of the early works. But, you know, all I want to say is it's two group of intuitive people that we meet each other. And then there's a strong synergy. And then fast forward, 2018, I officially joined as a board of director to their board.

And then I, you know, I have a better exposure of the company. And, you know, I understand they need to be extremely focused on their current roadmap. They have a lot of things to deliver. but I'm starting that a wider spectrum of recognition towards their industrial design in the past three, four years, which I'm super happy. And then Teenage also as a call funding design partner towards another company, Cal pays nothing. I'm not sure if you heard of a company called Nothing. They make phones. Oh, yes, the nothing phone, yes.

Yeah, yeah. We were behind nothing. So we also are power founders of nothing as teenage as a whole company. So we help nothing set up the initial whole design language and everything. And then when we started to do the R1, it's the opposite way we want to create something cool, which is quite funny because of course the first thing we're going to look at is who's out there, right? We saw some strong competitors, humane, AIPain, X Apple folks, huge respect to the team. But then I kind of like persuaded myself, I'm like, hey, you're offering this completely new generation of interaction software-wise.

And it's just to me, it's just too risky to offer a sci-fi gadget, which no one knows how to use it. So in my theory, hardware is never a choice by preference. You don't build a hardware because you want to build a cool hardware. Most of the time, if you do things like that, it will fail miserably. A lot of good cases I learned and I experienced is that you have a software that's so good that you want a dedicated hardware to make it better. It's always around the software. It's always around what's inside.

And then if you have a software that's very, very edgy, very new, you want to de-risk on a hardware, at least for the first generation. Got it. So to me, R1 is a result of choose hardware by necessity, not by preference. If I want to do hardware, I probably want to do a lot of other phone factor, like you said, right? How about a fancy glass? How about whatever that is? So the first thing we realize is that we want to set up a good position so that we can compete with all this big companies

and upcoming competitors, but as well as offering something that can resonate in your culture and in your memories and in your existing workflows that you don't necessarily need a menu to understand how to use it. It's hard to balance hiring top-tier developers and keeping your burn rate under control. But these days, I see a ton of founders successfully doing this by hiring remote talent. So let me tell you about Scalable Path. It's a software staffing company that can help you build an awesome remote developer team. And the right developer isn't just a list of technical skills.

We all know that. It's about their personality. It's about their work ethic, their motivation, and their fit within your team. And Scalable Path knows this. So here's what they do. Their team will get to know your vision. They're going to get to know your needs. And then they're going to develop technical challenges tailored to the roles you're hiring for.

and these challenges are conducted live and on video. So there's no gaming of the system. You're going to get great people. They also evaluate each candidate's soft skills like communication, attitude, and work style. Scalable Path has completed more than 300 projects for their clients, and they have a network of 30,000 developers. They've been doing this for over a decade.

They know what they're doing, so you're going to be in great hands. Here's the best part. Twist listeners get 20% off their first month. If you're ready to scale your depth, team and your business, check out ScalablePath.com slash twist. Once again, that domain name, scalablepath.com slash twist, 20% off. You've been using this product for a little while, you said, or at least the pie version. When do you feel yourself reaching for your iPhone?

When do you feel yourself reaching for the rabbit? When do you hit frustration points with either device? And how does that go into your thinking about, does this at some point replace me, bringing my phone with me. Yeah. And I have to maybe think about the hardware capacity of this device, or maybe there's, you know, this rabbit, but there's two other versions. Yeah.

Did you ever leave home without your iPhone and just take your rabbit? Talk to me about that. Yeah. Yeah. So first of all, any single direct actions, I'm leaning towards rabbit now. Why? Why? Because when I start trying to accomplish that task, it's a speed of thoughts.

and I have that already. And I feel like it almost quicker and faster, more intuitive than find that analog button without even looking at it and just talk. And the AI is accurate and good enough. Lam is fast enough to get things down. So I'll give you some quick reference. I'm always on multiple screens. I'm always on a lot of things daily when I'm working and maybe talking to other guys. If I have something that I don't know, this is definitely faster.

just forget about a lamp, just search-wise, this is way faster. Then I have to set up a new tab or go to the Chrome and start typing. Because you have the action button, it's kind of like a CB radio.

You press the button and you say, define the word iconoclastic or define the word Donnybrook. You don't know the word Donnybrook. You say define the word Donnybrook. I don't have to think, command tab, go to Google, define Donnybrook, etc.

Six or seven steps. Well, not even that. Some very advanced, you know, numbers, this gives you faster because I give you an actual case. I was having a meeting with one of our existing investors. They were asked about, you know, performance in comparison, you know, on sales and some numbers from other companies. I generally don't know. I don't know what the revenue from, you know, a company last year. And if you think about, I search on Google.

there's 200 tabs, which one is accurate, which one is correct. So that's why we set up the strategic partnership with practically just to enhance that part. So that's search alone. Another thing is music. I can tell you, you will love this just because it can play music. It's a second iteration of the classic iPod with the same level or even simpler level of control of this. I had my own privilege to have a phone call with Tony Fidel actually yesterday. So we were on the call for more than three hours.

And I think just music play alone, that's definitely my number one in the past eight months. Other than search, search is probably 70% of the music. But I think a lot of people still don't get it. I start posting stuff on my Twitter. There's one scenario. I use the vision. So basically, I double-tap the camera rotates and points to whatever you want to point. I use the vision to look at the Discord.

Because Rabbit now, in the past three days, we have 5,000 members. And I'm the customer service guy there. I want to listen and learn from my own. Always good as a founder to listen to the customer support line. That's where the truth is. I reply almost as much as I can, but I start getting lost, right? Because the message is too much. And I was actually on a Zoom call with some other guy.

And then I literally just point in the camera and said, what are people talking here? Oh, a lot of people don't get them. A lot of people are thinking, oh, you have eyes, you are not blind, and this is stupid, and why are you doing this? But I'm actually on something else. And I don't have time to scroll my mouse for 50 pages to understand. Even though the current generation of GP4 Vision and our own vision model needs to be faster,

actually it needs to be 10 times faster by shipping, which I'm pushing it. But after four or five seconds, it says, oh, here's a conclusion. People are talking about how Rapid R1 can potentially work. It gives me the report. Which you would normally, you might have a college educated person who you say summarize what's happening in the customer support line for me every day. Or you might point an LLM at it. and have it do a report.

Here you just take a quick picture. Yeah. I, too, was having this problem. I was skiing this weekend, and I kept trying to change my music using Siri. Right. And it is so painful to just lower the volume and or change the track or play a different playlist. You're talking about like five to 10 seconds of Siri getting it wrong and then you've got to

do it again. And then your music cuts out. The idea that you can press this button and then very quickly, you know, the response time. I think all of those smart speaker systems, which the Raven, you mentioned a couple of times that you sold to Baidu, that was your first company, I think, or second company.

Yeah, first company. You must have learned a lot about just response time to voice commands. It's just so terrible to wait that four or five seconds for it to wake up and to understand what you're doing. Right. And that's something you're trying to solve is that first five,

what do you call that, the five second delay? Yeah. So, so right now, If you break it down, the delay came from two parts. The large action model, again, is fast. If you go check the demo I released on the Twitter, it's fast. Play a song is instant. If you ask some random generic questions, we have a technology just basically make the streaming tokens with LLM make it very fast.

If you ask what's the nature of reality, what's the difference between orange and the mandarin oranges, anything that does not require up-to-date information where a 500 milliseconds. Anything search up-to-date information, that starts to slow down. If you try open AI in the first beginning, I haven't tried recently. I think recently improved again, but usually we're talking about two to three seconds. Instead, in comparison of 500 milliseconds. But I figure the visual part is where the most delay came from. We're talking about somewhere around, you know, 8 to 10 seconds. But that's not us. That's the fastest in the industry right now. and we're working restlessly trying to figure out a way to further cut it down.

But really, right now, the delay is really just search up-to-date information and vision. That's it. If you just trigger lamb stuff, that's fast. That's no-brainer. That's very, very fast. But frustrations, let me tell you the frustrations. So, yes, number one, I don't want to carry two devices. No one wants to carry two devices.

I'm not trying to convince people that this is just sexy so that, you know, maybe you want to carry it. No. What we can do for now, we know it cannot be an app. So same footprint as your iPhone is actually very, very light. It's only 110 grams. So what's 110 grams? Go to your fridge, grab two eggs on cook eggs. That's it.

And it's light enough so that every single time I put in my pocket, I forgot about it. And then I appreciate myself for the design of the analog push to talk button because I don't really need to take it out and look at the screen. I reach out to my pocket and I feel that button. That's actually how I use it most of the time. I just connect to my AirPods

or whatever Bluetooth devices or car systems. And I don't really need to look at it. I reach and feel. But still, there's a lot of things I would rather go to the phone, at least for now. First of all, heavy social features. This was not designed, at least for this generation,

This was not designed for connecting all your friends and chitchad and figure out what's going on. This is more focused on task solve. So that part, unfortunately, I have to go back to the phone. Another part is- For your group chats. If you want to get in your group chats on Signal, WhatsApp, whatever your jam is, I message, it's not going to replace that just yet. But you could.

We could. I mean, it got a tray. It got the same card. Yeah. It is a phone. I mean, come on. It is a phone. We're not trying to be a phone,

but it is capable of doing all the features of your phone. Another thing is, which is, I'm not sure if it's just the fact or if it's just my angle, but phone is actually a content-consuming device, right? Like if you think about highest gross apps on the app store right now, it's Netflix and stuff. It's TikTok, Netflix, Instagram, and all that stuff.

That, you know, I mean, come on, iPhone has a better screen. Yeah. Big screen. Yeah. Big screen. Better, much better screen. I have to say that. That's why, you know, when I start thinking about laser projectors, I'm like,

I'm not sure, you know, like, because I showed it, I saw, I saw the humane demo that they do the gestures to send a message. I'm like, no, I probably wouldn't do that here. You know, I. No, it seems crazy, like forcing it. Right now, startups have to do more with less. We all know that. And that means increasing your product velocity while maintaining or even lowering your

costs. Now, don't forget, product velocity is how startups beat incumbents. So here's the great news. AI is going to help you do that. So let me tell you about Wizard. It's spelled UIS, A-R-D. It's an AI-powered suite of U-I and U-X design tools. With Wizard, you can generate your app or web designs from simple text prompts. You can then iterate on these designs with an AI assistant, and then you hand off your completed designs as React or CSS code. Wizard's text to UI mock-up tool is called Auto Designer, and it's really cool. If you're watching, you can see it on the screen right now. Here's the brass hacks. Wizard is going to help you go from idea to mock up in minutes. So if you're creating a product from scratch, this is going to save you so much time.

Start building products today faster with 25% off Wizard Pro at wizard.io slash twist. That's u-I-Z-A-R-D dot I-O-slash-T-W for 25% off. Stop wasting time and start shipping faster. the voice recognition is getting better and better. And these LLMs seem to correct errors in real time when it's transcribing. So I'll talk a little bit about that, the ability to transcribe and translate voice to text really accurately and how that's changed things. Because when you were running Raven, the speaker, it was kind of janky, right? Wasn't it terrible voice to text? Yeah.

When I started Raven, that was 2013. I was actually working with SRI, a couple of folks in SRI, Stanford Research Institution. That institution later started a nuance and nuanced incubated theory. So it's a long history. But I remember very clearly the first time I paid a visit to SRI, the dictation was like roughly at 73, 74% native English, which is very, very bad. But it went up very, very quickly. Another problem is not the dictation. The problem is intended understanding.

And then again, we're talking about pre-transformer era. And then back at the ribbon, we worked really hard on natural language processing or NLP. That was the best technology before GPT can understand you. I guess the idea is not because we didn't thought of we could design an algorithm in such way. It's just back then we don't have strong enough computation power like GPUs, like an Nvidia GPUs to be able to run that. So if you talk about intent recognition, it's really, really painful. It's heavily human consuming because essentially all these smart speakers that I created, Alex and every previous generation, when we first started creating that,

it's like you're assembling a menu to this quote-unquote AI, and you're basically hard coding that would be like, hey, here's 70 ways that you talk to the speaker. That means you want to listen to this track. I want to listen to, play to, get it on. You know, like there's a cluster of sentences you can describe what's the same intention. And NLP is for us to figure out that, to assemble that menu.

So a lot of things were really, really not ideal. And I totally get a skepticism towards R1. Be like, hey, it's a voice first device. I saw that. That's no goal. I get it. Because we are all the same. same generation of consumer that still share this PTSD from early days of the very poorly.

Exactly. Siri has been so frustrating for people. And Alexa was better but not available on your phone. To the same extent, Google was assistant never really caught on. And so we just assume it sucks. And the truth is, it's much better now. Even using the chat GPT4 app. Yeah.

I took my, you know, the iPhone 15 now has the. action button, which is similar to what you're doing on Rabbit, I connected that to chat GPT4's voice interface. Right. So I actually, when I press that, it goes to that, what are they conversational, I guess, mode in the app. And it's quite nice. It's a little bit slow.

It's like the hardware seems to be working against it. Yeah. Whereas in U.R. It's going to work quickly. When do you know you'll have made a product that is ready to challenge the phone in your pocket? and you will win the which device I go back home for. If you forget your phone route, you go back for it.

If you forget your wallet, you don't go back for it. You're like, I've got my wallet. I got some payment options on my phone. When will your device, you know, if you leave home without it, you got to go back and get it. When will you hit that milestone of I got to turn the car around and go back and get my rabbit? Yeah.

First of all, it's such an ambitious task to challenge, I'd say. We never set up self-exam. at least mentality-wise, we never set it up ourselves to be like, oh, the whole purpose of rabbit is to do this, it's to kill iPhone, whatever that is. But on the other hand, we understand deeply, especially given by my experience, you know, my company sold to Bidu,

and we were at a very exclusive working relationship with that Raven still keeps driven, but I have seen enough, not only Bidu by Microsoft, all the other companies, how big company works, you know. I think we're at a point that. we don't think we're confident to challenge right now, but we don't want to wait either. Like, yes, you can be a user, right? I post this on Twitter.

I said, okay, would you rather be the wave itself or the best server in town or that floating body in middle of those? Right? Because you can only be three of those things. So I think what we are is that we're not saying, oh, we are so ambitious and so dilutional at the same time that I have a very clear master plan, this is going to work.

That's just not the nature of startup. I mean, you've seen enough of startup things. You never think like that way. You're like, okay, I want to do it. I want to do it now. That's how it works. But I don't think that one thing I have strong confidence is that for all these current generation

of application-based operating system, it's impossible for them to improve. Because technology is not, It's always not, I mean, fundamental operating OS shifts. It's never be an improvement. It's always reramp. It's always revolution. And the problem is not about technology.

I think a lot of engineers having a different opinion towards engineer perspective, it would be like, oh, you know what? They just do this and that and say it becomes R1. And they become rabbit OS. It's not like that. The problem is they're incentives behind how they make money, you know. They set up this entire application store. with billions of developers, billions of apps,

and all of a sudden they said, you know what, guys, no apps? How does that even work? I don't see a very smooth transition how that can happen so quickly to the point that we think we should wait a bit, and further lower our risk. We cannot do that.

We just start right away. But I think, consider R1 is iPod, you know, it's a good phone factor for a certain niche amount of We're talking about a slightly wider request than Apple because iPod really just to replace Walkman, right? But in the iPod era, you still have your BlackBerry phone. I do that. I have my BlackBerry 800, 800, I have my iPod.

Two back pockets. You put one in each back pocket, right? Exactly. And it kind of like fixing the shape, you know, especially if you're wearing jeans. You don't even want to swat. You just, you know, keep it there. But, you know, we have R1. this is our first attempt or approach towards a future structure, you know, the future,

how software will work with human, how you interact with software. But I think maybe in next a year and a half and two years we'll have a better answer. Yeah. And for now, you're going to go after, you know, people who are early adopters, the avant-garde, people want to try it, people who want to invest in the time, and give you really good ideas. Have any of the app developers come to you directly and said, can I just make stuff directly for your lamb,

your large action model? Can we have an interface where you have your language action model, you expose us to it, and Uber engineers and your engineers can be in the same space, you know, collaboration space, making it operate flawlessly. It's actually happening from both sides. Jason. It's one side, yes, you know, you saw the

Propractic deal. Yeah, that's a new way. We literally handshake on Twitter. We have no pre-communication, you know, we just decided to do it. And now we offer the first 100,000 units with $200 credit of propractic. If you think about the device itself, it's 199. So it's happening from this side, company to company. But more importantly, you know, I think the most underrated feature of Rabbit OS, which is kind of

experimental right now, but we're confident to roll it out, at least for the beta version, where soon after people start receiving their advice is a teach mode. And how the teach model works is that you actually explain that before, is that as if monkey monkey, like you go to this white portal, right, the rabbit hole, you turn on teach mode, you decide today I'm going to teach you something. And it's like if you're teaching your eight-year-old children, you'll be like, hey, come to daddy, sit right here and watch exactly what. I'm doing, and learn from it.

So if you think about neurosymbolic, same metaphor. And there's a psychological factor behind it is that, Jason, if I ask you to teach me Ollie on skateboard, would you rather do that or not? Yeah. There's two decisions. It's one, you want to do it, or even you want to do it, you probably wouldn't wait into. But more importantly, if you don't know how to do Ali at all, you would just say no.

Right. Correct. So you go to the. teach mode as user, you know exactly you're going to teach. Right. You're not going to teach how to use a skateboard if you've never been on it or skiing. But if you said, hey, how do I carve skiing? And if I opt into that, it's going to be good training data is your point.

Exactly. So if there's an app that people aren't using, like, let's say you haven't trained Twitter yet, I could say, hey, let me train Twitter of I like to go to my replies. I like to go to my mentions. And I like to read these out loud. Yeah. Please read me five of them very quickly. but I don't need you to read me the date.

I just want you to read the content of it and just read me the first 200 characters, not the whole thing. Exactly. You are already using the teach mode. That's the powerfulness of it. So the psychological factor is you go there, teach under the impression of,

you know exactly what you're doing. So the data you connect. Of course, we don't record on your local device. We bootstrap the virtual machine. And the virtual machine is like old technology. So you go to virtual machine, choose whatever platform you want to teach on, and you go teach.

And that recording is very, very clean. So that the Rabbit OS can get one shot in just one shot. So you're letting the entire community teach the Rabbit OS. To build your own rabbits. Amazing. That's how we make money, by the way. Answer your question. Got it.

So I could make a rabbit, which would be my way of interpreting, say, TikTok or social media app like we talked about. And I could sell those. I could sell my own rabbits to other users. Yeah. So we're not reinventing any business model. It's classic Apple App Store or Google Play Store logic. But we're not worried about making money for now because making money... Well, you just raise $10 million from Vinod. Congrats on that. He believes in it. And how big is the team? Like, how big of a team do you need to make a hardware device like this and build this stuff out? Because we see a lot of people doing more with less. I'm curious. Yeah. So we had a full headcount roughly around 17 people to. date, but we're hiring, but I'm also working with good ODM, OEM, I had resources over the past Raven career.

So it's the same team, plus, you know, design armate Teen H helps us really a lot. But we're hiring, yeah. We're working on the future version of LAM. So consider this is LAM one. We have a LAM2 in design, which that's going to be much more interesting next year with with different phone factors. But the general idea is that I have worked from my past companies that I know that if you're aiming to make money on day one, you're probably not going to be a consumer

hardware company anyway. You become more like a, how do I put it, like Oracle. Well, you seem to have been very inspired by the iPod. Yes. And if you think about the iPod, my favorite iPod was. the clip on. I don't know if you remember the second generation of the shuffle. Oh, I mean, that was, I believe, like peak Apple insanity. It was a clip. Yes. It was almost like a tie clip and you could use it as a tie clip. With a shorter airbus,

with a shorter headphone cable. Yes. They have to change that too. It's a loop. Yeah, I remember that. It was a short headphone cable because they knew you were going to put it above your waist. You didn't need to like run it to your pocket. But it also had physical controls on it. So it was very elegant in that way, but it, you know, it was for other people who love to have a large collection, you know, the classic iPod, the fat one, you know, the original iPods, you know, those were amazing just in terms of having a flywheel on it with that physicalness to it, right? Then you have that scroll wheel. What does the scroll wheel do on this exactly? We're working on different ideas that I couldn't share much, but I can share one idea if we're working on. Yeah. This is how you're going to, this is how you're going to unlock your device.

Ah. So your passcode will be like a lock when you were in high school. Two turns to the left, two turns to the right. Ah, that's super clever. With haptic clips on it. With haptic clips on it. Ah, so I go one, two, three, push in. Down one, two, push in, up, push in, boom, three, two, one.

Whatever it is. Yeah. That's super clever. And no one's ever going to know that, right? Because you don't, it's not even, it's not even a text. or anything. It's just bad. Yeah. And so there could be different sizes of this. You said, like, you don't actually need the screen.

Do you need to make your own responsive headphones or use your Bluetooth? Because one of the things I have a problem with is, you know, I have my AirPod third generation. It works so elegantly. But I prefer the pixel buds. I don't know if you use pixel buds, but they're flush flat. They're a much more elegant device. They work better with a ski helmet.

Work better if you're sleeping and you want to listen to a podcast before you go to better and audio. like I do. How do you think about that? Because if it was a unique headphone, man, would there be hardware gains or user interface gains that you can achieve? So definitely, if you if you customize from kernel perspective, you can further lower down the latency. The latency is an issue, especially if you try in car system, you know, that's why all the car series doesn't work because the car system latency is very well outdated. So it has a very long latency. But right now it's Bluetooth 5.0, which is by default work with any different default AirPods

and third-party manufacturer. I'm more interested in maybe we can do a rabbit AirPods, but it's fully standalone. Hmm. That's fascinating. So you put them in and it's your version of the rabbit. I just talked to it and I authenticated it. So I just leave home with my rabbit AirPods.

Yeah. And say, hey, order me an Uber. It says, okay, Uber Black or Uber X. Right. You say, Uber X, I'm alone or Uber, you know, SUV. And it's okay, yeah, boom, we have it. Oh, wow, that's sick. And that, that challenges came from, one, you're waiting for a smaller chip with more power

because it's little room to wiggle with. You know, if I do an AirPods max, I can do it now. I can do it now. But if you want to do it, right, there's some challenges on a chip. So the base station, your, your, your, your, your, would have to have a SIM card or ESIM or whatever. Battery life starts to come into play here. Yeah, a lot of different issues.

I mean, that's a sick idea, though. If you had over-the-ear headphones and the rabbit was built into it, then you just have those. And oh, my Lord, that'd be incredible. Has anybody ever put a CPU inside of headphones? I've never seen that. That would be amazing. There must be some in Che-Shan or Akihabra.

We could go find. Syriatically, I mean, Shenjin, I already checked. You know, I called my buddy. You checked. I checked. There's, there's, yeah, there's, there's, there's, there's, there's, there's, there's, there's Apple. There's two big, different parties are working on that.

We're not only talking about. Oh, so they might come out with it, you think? They might come out with, uh, I think, I think that's where I was working towards two, but. So I would put on my, just, what it's clear, I put on my Apple Air Max, whatever they call those over the year ones. Yeah. And then I say, hey, Siri, play me the clash.

And I don't need to have my phone with me. got Wi-Fi and LTE belt into it, that would be amazing. And Jason, I'm going to show you something. This is the actual board of R1. It just that bit. Oh, okay. So it's the size of- We didn't even try to make it smaller.

Because look at the phone factor, it fits, right? It fits easily. We didn't even try to make it smaller. If I tried really hard to make it smaller, I mean, definitely can. I mean, it's half of a playing card. If you cut the ace of spades in half, that's about how big the motherboard is. as it were.

Yeah. So over the air, I think we can probably do now. But I want to do the AirPod version. Yeah. That's going to be a little bit of challenge. And also because, you know, you want to set it up to always listening. And then the battery comes the issue, you know, so. Always listening takes, yeah, a little bit more battery.

Yeah. Yeah. So, but I think, you know, I'm pretty confident in the next, you know, 48 months. That's something that we can start figuring out. But another thing I think is very, very good to think about. And don't call me on this. I'm not leaking whatever. This is just my thinking logic.

You know, like, I'm not trying to say, oh, this is what we're going to do next. This isn't the product roadmap. We're just riffing here. This isn't. Yeah, this isn't. I think a watch makes a lot of sense, too. A lot of community members already designed Rabbit Watch on Twitter and tag me. Amazing work.

But if you think about it, Apple Watch is probably the week. biggest product in their entire product by animal. I have to say Apple Watch sucked until the Ultra. Yes. The Ultra. Battery. It's better. The battery is better.

The screen size is better. Yeah. And I like when I jump in the water. Yeah. That it tells me how many feet I'm underwater. Right. Right. It's just having the battery life actually the last two days.

Yeah. It's such a game changer compared to the other one. And the screen being, I don't know, What is the screen? 50% bigger than the other ones? It's 25%. It's very bright. It's very bright.

And that's a big factor, too. I mean, the way I put it in this from my own experience that I said, maybe Apple Watch is the weakest in their entire product line. It's that it's an accessory. It's not a standalone device. They really haven't focused on it as a standalone device. I agree. They said it's a standalone device. I understand you can make phone calls.

I understand someone. It sucks for phone calls. Some YouTubers carry that for a week. I saw the YouTuber challenge. Oh, really? Yes. I think someone really, like, have their Apple Watch for, like, a week. It works, but it's not designed for it.

I think there's a good opportunity or ideas in general that, because I start paying attention to kids. What is that? I start to pay attention to how future generation are kids, how kids are using their devices. Oh yeah, this is one of the community designers. I just wanted to show this while we're on air. This is a fan made this. So for people who are watching, it's the iconic orange, and it is gorgeous.

And then it's got like, you know, the crown. But the best part about it is it has a pop-up camera, and the screen is designed with a little carve out of a pop-up camera. Completely impractical, but awesome. Yeah, it has to preserve all the key elements, right? like the venting grills and all that. It's super impressive. But how did this community start, by the way? I just have to stop there for a second because did you goose this community of lunatics in industrial design or is it they follow teenage, what is it called teenage engineering engineering?

Engineering. Is it that teenage engineering inspires other engineering students or people? It's nuts to watch how many people have started to make derivative products out of your iconic. Yeah. And it is iconic. I know you don't want to say that, but it's such an iconic device that it's inspiring people

to make very refined product extensions for you for free. Yeah, exactly. So I think, first of all, huge credit to teenage engineering. If you ever, ever worked in design or industry design or music for one day, they're a god tier. They're like the ultimate outer.

At least that's my opinion. I don't care if anyone agrees or disgrace to me. They're in my heart. part, the gut tier. So that's one part. Another part is Jason. Phones are just too boring now. Industrial design is pretty much that.

Explain that. I can't tell the difference between my last five iPhones. It's all the same today. Exactly. Like phone companies, they set on their current phone factor. Let's think about a fun time, right? Let's think about sidekicks, engage. I think about those things,

like flipping Nokia, like spinning Nokia's. I think it's been so long if you look at the past, especially past five years, I think a very good example from Samsung and Apple. You know, now it is iconic.

I saw an iPhone, I know it's an iPhone. I saw a galaxy, I know it's a galaxy phone. Yes. But I also can tell you 100%. The next generation will be looked like the same. Absolutely.

With a new difference. different car. And this is because of manufacturing. These things are so expensive to start a new version. It's exactly like cars. Yes. Once you make the Toyota Prius, every time you redesign it, you're basically restarting like a whole production line and you don't get any of the scale.

So it's a scale business. They're trying to get massive margin in a competitive space. So they don't change it. Yeah. But I guess Steve Jobs doesn't do a f***. he does not give a f*** you can be sure he did not give a it just it just that the spirit is generally dead and I think we're not helping we're not we're not we're not basically

doing this it's just it's a common feeling for us all of us as consumers that that's that if you bring something new and fresh yes it is it is relatively cheap yes we're a new company but people see the joy of it by looking at it and then then that inspired them to create their own iteration of it, which is another round of ripple. And this ripple just started getting bigger and bigger. And I really, really understand why Kyle Pay started nothing.

In the beginning, we have a couple chats. You know, his whole company's position is phones are boring. That's it. That's why there's nothing. I mean, the nothing phone is wild. What do you love about the nothing phone? What is it that inspires you about it? I think they are probably the most engineering and industrial design invested company, you know, giving by the fact they're still a startup.

Most of the startup wouldn't invest this much into both of these fundamental areas to make a product. I think I give you, I gave them huge credit for that. Yeah. And, you know, I haven't developed directly, you know, involved in their management or anything like that. But, you know, being with Teenage, seeing this company from beginning, it's quite a journey. It's quite a journey. It's very challenging. You know, their stuff isn't cheap.

You can tell it isn't cheap. It is not cheap. It's a luxury item. And they're trying to make, it is a luxury item or not? well, it's still a little bit cheaper, $100 cheaper than the flagship Android phones, you know, I'll say. But it's not as cheap as like Shenzhen stuff. Yeah. So I guess, you know, the younger generation, if you think about younger generation, like all the kids that are born after 2000, they, when they grow up, have the first access to the phone.

and then for the past five years it's the same boring stuff. How crazy is that? And think about when we started to have our phones, we have 200 different things to choose from, right? And we're getting a phone just because it's cool. That's it. I think that part is mostly gone.

And you're right. It's because of what's behind the manufacturer logistic stuff. But I think it's good for us because we're so small and so in the beginning that we can take this this massive risk yeah you can take the risk we can take this

but also you have nothing to lose you know nothing to lose like if you make this form factor and it sucks you change it and you try the next one and you try the next one until audience responds if Apple screwed up an iPhone 15

the stock would lose $500 billion in fact when they've had problems shipping problems etc something gets screwed up they lose 100 billion dollars in market cap and then everybody loses their mind. I mean, this is one of the great things about being small.

Yeah. And that's exactly the reason why I don't think Syria and Google assistant or Alexa can catch up so quickly. I think you have to throw it away. I mean, honestly, I think the technology has moved so fast that Siri and Alexa need to be thrown in the garbage and they should just start a new project. It would actually go faster. Yes, yes, yes. I agree with that.

The technological debt of Siri must be insane. The entire paradigm has shifted, right, to lambs and LLMs. And they're not built on that. Yeah, I mean, it's a 10-year-old product, you know, so. Toss it in the garbage. I mean, that's, I mean, this is what, you know, if you look at what Elon's doing with self-driving. Yeah. I think he doesn't have that fear of saying, hey, we built this software to stay in the lane,

but it's not, you know, there's new technology that's going to disrupt it. We need to move to that. Now, the hardware stack may not change, right? The hardware stack could be the exact same for the iPhone. Yeah. But Siri has to die. You know, one thing I want to give huge credit to Elon is that, you know, there's certain ways he thinks that I think really, really resonance with me. One example is that if you look at all the cables in a cyber truck, I'm not sure if you order one.

I have one coming in. Yeah. If you look at how they redesign the tables, it's crazy. Because all the cars have like what like now, like 10 miles long of cables all around. all wrapping in different payers. And he just doesn't like that. He said, okay, that's too complicated. Let's make one big fat cable.

And with two plugs and that's it. I think this is some thoughts that really pushes a lot of innovations. And from where we are benefiting from it is that, like you said, we're small, we take the risk. But this is how we want to think about things, you know. Like we're not saying, oh, maybe we wait for. for two more years, there's more API we can work with. Maybe we can, you know, suit up and walk into four apps offices, talking to their executives and maybe we can get a deal.

We're like, okay, how about we get rid of that? We don't work with APIs. We don't work with apps. We just do something from grown up. I think we're hugely benefit from thinking like that. Yeah, the cyber truck went to this 48-volt system, which reduced, the total wiring was reduced like, I think it was 80% in the car.

Yeah. It's bonkers. And, you know, now, I guess the only downside to that is if you lose that cable, yeah, the car is done, right? You lose so much. But then the upside to that is, if there is a problem, it's easier to diagnose you swap the cable out. Yeah.

I mean, it's hard to lose the table because you got bulletproof panels, right? Yeah, exactly. You can't, I mean, Joe Rogan tried to hit it with an arrow and it didn't work. And he's a pretty good shot. Yeah. So you're good. You're good. Listen, Jesse, you're like an amazing guest.

I want you to come back when it ships. When are people going to get these in their hands, do you think? What's the best estimate? Okay. So we will have our promise date is Easter, which is March 31st. That's where we start shipping. We're working with local compliance and legal teams to get international orders ship as fast as we can.

Easter seems reasonable. Easter seems reason. Knowing that we said, okay, the international batches should be, you know, could be June, July, but we're trying to move everything faster than it should be. But we can allocate maybe up to 50, 50 to 100 somewhere in that range of media test devices. Great. Thank you for including me. I'm super psych.

Yeah, you already signed yourself up. I know that. So that's expecting somewhere mid-October. So in the next two, three weeks. and I will start rolling out all the demos as I showed on a keynote starting from next week I already showed a couple bunch

here and there but like I said latency is something that we want to really really focus on and How many years have you been working on this by the way? I think this is something people don't appreciate it that this is not something you've been working on for six months no the original idea like I said Rabbit is basically

Raven 2.0. It carries a lot of the it's the same original passion and carries a lot of the legacy But if you want to meet, won't have me to date, you know, backtrack to day one. I actually had my YC application video that community members put up from my like YouTube channel. And that was dated as nine years ago. Nine years ago when you started this journey, two companies ago.

And then this company, you formed a couple of years ago, right? 2020 or something? Yeah. It's just, it just, you know, the idea was always right. I guess the direction was always correct. But unfortunately, like I said, 10 years ago, we don't have all these cars on the table. Like, the intention wasn't solved.

There isn't GPU files to utilize at this current rate. So I feel like I just waited until, you know. The hardware, the infrastructure had to catch up to the vision. And that's basically it. Exactly. And yeah, I will say that, you know, it's been sticking with my hand for so, so long. And that also explains Jason that we're moving so fast. We're not starting from nothing in two months and build a hardware.

That's impossible. Right. Yeah. The ability to create a hardware startup today is much different than when you started 10 years ago. Exactly. Yeah. I mean, the sophistication level in Shenzhen as just one example. When I went to Shenzhen for the first time, I think it was 2006 or seven.

Okay. It was It was so rare to see an American, a white person that when I went there, I kid you not, 20 school students surrounded me to touch me and take a picture with me

because they'd never seen because they had come down from the north. There's a famous amusement park in Chechen and it's not for American tourists, it's for Chinese, you know, native nationals. And I went to it just to check it out and I literally got stopped,

you know, 10 times. Just people wanted to touch me and take a picture. Now there's a lot of Americans there. There's a lot of startups there. And literally when I was there in 2007, it's 15 years ago, I guess, there were more, there were, maybe like a tall building, 30-story building. Yeah, yeah. And for four blocks around it were all parking lots.

Yeah. And you would drive from one building to another. You could see the building you were driving to, but it was like driving through the desert. And you know what? You see that in the Middle East now. Some of the regions there were like, boom, a tall building. comes up, boom, another draw building comes up, and in between it, there's no infrastructure, just like a road, but nothing else. It was a crate, now you go over there, my lord.

I left my, I was originally born in China, but I was raised on, you know, we're a Western-style family. You know, like, my parents be like, once you're 15, you're on your own. I started learning English right after I. Where did you grow up? I grew up in Shian, where the terra-cotta warrior is, but I, oh, yeah, but my parents start teaching me English on day one. And then they're like, once you're 15, you're on your own. I literally left my house at the age of 15. I got a transfer opportunity briefly started at a high school in Singapore.

And then I went to UK. So I graduated from University of Liverpool mathematics. And then spent another year and a half in London. And just to hang out with a whole bunch of artists, U.S. So I don't really understand China now. You know, I've been living the country so long, but I believe you, like, you know, it's crazy. And the heavy ODM, OEM, OEM infrastructure that they set it up over the past decades is still, you know, at least serving Apple. You know, all the iPhones were made there, you know.

Yeah. And, but, you know, I think a lot of things are starting shifting a bit. bit because I think what I can see from hardware perspective is that a lot of things are become modular, but on the other hand, core chips are become centralized. So the core chips are being centralized. TSWC, yeah, TSMC. Taiwan Semiconductor. Yes.

And Invidia are locking that down or, you know, Arm and other folks are locking it down. Yeah. But the components, anybody can make them and just plug them into those CPUs and GPS. Anybody can make it. But I think no one still beats the Shenzhen speed. That's a cool part. That's a cool part. They can fabricate something for you that's brand new in days and have it back to your office in America.

Exactly. I mean, the iteration cycle used to be months. Now it's weeks or days? Now it's definitely weeks. and if you have some special resources, you can make it happening days for sure. Which is wild when you think about it. It used to be like quarterly you could

get your product, iterated on. I know because I have a bunch of hardware startups. I tried to invest in 10 years ago and it was just the iteration process was too slow. And you couldn't be nimble. Now you can get it done in weeks, maybe even under a week. That's extraordinary. Right. Yeah. But I see

the chips are getting centralized, which is not very good for startup, to be honest. Why? Oh, because, you know, you're going to work with only selected partners. And then they have so many big clients to work with. And then they're like, okay, if you don't sell more than this number, we don't even talk to you. Got it. So, yeah, if you want to get the attention of Nvidia in the age where Facebook's buying 10,000, you know, H-100s is going to be hard.

What do you think about risk and the open source chip movement? Is that going to have an impact at some point in the future? Risk 4 and stuff? I think it's more from how you want to design a system. A lot of people are debating. You know, I've been debating myself and been debating with my own community on Discord. A lot of people are like, hey, can we make a local LOM? Can we run X amount of features on device instead of in the cloud?

Yeah. For me, I get it from the privacy reasons. I get it. I think that's probably the only valid reasons. But why are you even willing to run it on local? because... Yeah, and how many people care about that?

Very small number of people. If you encrypt my instructions, do them in the cloud and then delete them. Yeah. What's the... I mean, or you delete them every whatever,

60 days or something. You may need to have them for some historical purposes, but... It just, there's a huge dispute between academic perspective and consumer perspective, right?

Consumers have voted. They don't care. They don't care. Exactly. But I think, you know, the way I bring this case is that there is one way that we can work further towards like Chromebookways, right?

Like even the operating system is in the cloud. If we go that route, then we don't worry about chips because you don't need that chip to run locally to heavily do the tasks. It's a terminal, right? You turn the device into a down terminal. It's a terminal. I'm actually pro that route, you know,

just because from consumer perspective, why the heck do I need a CPU in my car leader box? Yeah. Why the heck do I need that? And I think, you know, everything is smart. The TV is smart and everything is smart. I actually, I was setting up a little apartment because I'm getting too busy. I stayed out of office.

I was setting up the TV. And then I need to do the airplay thing from my content of computer. So I went by, I bought a Apple TV, and then I bought a smart TV from Samsung. and they both have Netflix. Yes, and so now you have two remotes, and one of the, the Samsung remote has the Netflix logo on it. It's like, why did I buy the Apple TV?

Exactly. And then Samsung had their own way of Chromecast or whatever. Yes. I think, you know, then if you think about what's happened behind, is that do we need all that? And trace further down, do we need all that chips? Do we really need a Qualcomm in your TV, if your TV can just project things?

and that's the TV should do, right? So I think, you know, it's a funny, it's an interesting trend to watch, especially in the next couple of five years. I always see who can get upper hand, you know, the local guy versus cloud guy and how they works. But yeah, I do agree with Elon Musk and a couple of other folks that we will have a power shortage if the AI runs like that. Oh, we'll have a shortage of actual power to homes, to data centers. We're going to need much more power because these things run hot. Yes. And also, we don't really have like transistors and transformers to be able to handle the cluster of the GPU if we keep the current pace.

So a lot of basic infrastructure to look at. But for us, you know, at least for now. Now this will get people interested in nuclear again. If we start having an energy crisis because of so many H-100s and then eventually 200s, whatever being deployed. People are going to say, you know what, we just need to start building some nuclear reactors next to these data centers and putting in massive solar farms and really having much more, much cheaper energy.

Right. Which, you know, people haven't had to think about it all that much because we've been able to keep up. But if all of a sudden we can't solve cancer because we don't have enough energy, we have enough chips to cure cancer. Yeah. We have that software. We have the chips.

We don't have enough energy to put at this because of custom medicine or something, you know? For me, it's how far away, if you're a consumer, how far away from the actual problem, right? If you're very, very far away, you don't care. But if you're next to it, and then all of a sudden become a problem. So I think, you know, like I said, we are, we don't know whether this is the correct phone factor. You know, we don't know. But we know this is the lowest risk we can pull off. and we know the lamb is most important.

Yeah. I mean, the lamb is like a sniper shot. You can just very easily think, accomplish the task. And, you know,

when we started in the industry, I remember in the 90s, there was a company called General Magic. Oh, and they were making a device that had agents. And you would tell the agent, I want to fly to Paris.

Yeah. And I mean, it would know you're in New York and it would go out, try to find tickets for you, go to literally go to a human travel agent and then try and get

the results back and then put them into some format that you could say, okay, I like that itinerary. And now here we are, 2005, 2015, 2025, yeah, it took 30 years basically to have that work. Funny enough, General Magic was a

internal Apple project that Apple don't really understand and spin it off. And their logo is a rabbit in a hat. Ah, right, I remember that. And the rabbit would pop its head out in this little animation. This was, I remember Sony made the first actual hardware for general magic.

And it was, some early iteration of the PDAs and stuff. Yeah. Yeah. Personal digital assistance. They were, I remember when the Palm came out in New York. Right. Right around 97, 98, 99. People had a StarTac phone and they would have a Palm.

You take these two devices out. And you learned Palm script. You learned how to do script on the touchscreen. And the number one application wasn't sending money. That was like what Elon and Peter Thiel were going to do, or Peter Thiel was going to do with PayPal, it was to send a business card. And then what's the latest feature of iPhone 15 is tapping the two phones together to send a business card?

Yeah. It's hilarious. Yeah. Same stuff. That's actually really, really good intake because I deeply believe that, you know, human are never inventing new demands. We have a certain category of demands, right?

We want to rest well, we want to eat while and drink well and be entertained and, you know, be motivated and start work. But technology is not. reinventing new demands. Technology is solving the same problem over and over again, each generation in a much more intuitive way.

Yes. And that's exactly what you said. With the palm sending that scripts to name card and verse it now tap to error drops. It's the same thing. It's just much, much more intuitive. And hardware is very interesting because if the hardware is intuitive enough, then the phone

factor stays for a very, very long time. I never I never heard another two thousand companies every year claiming to make a better hammer. No. Why? Because it's intuitive enough. It only takes your force and obey the phases gone. That's it.

Intuitive like that. But then if you think about computers and mobile devices, there's long, long way to go. Long way to go, yeah. Long way to go. And that's why we decided to take a little step forward. We're like, okay, I understand you guys. our skeptical about this and that, but we have to try.

We have to do this at least, you know, like, there's something about the instant nature of this that I think, you know, if you save people, I've always felt when I look at startups to invest in, is this startup saving people time? Yeah. Is it saving people money or is it entertaining them in some way, right?

Right, right. And so if you look at your device, it's going to save people time. And time is the equivalent of money for some people. Some people have more money than time. Some people have more time than money. But generally speaking, I think yours is a time-saving device. And then eventually, will it be entertaining?

It could be, sure. There could be entertainment services on it. But just saving people time and letting them get back to their kids or to dinner or whatever and get off their device. That is extraordinary. And so I think your recognition that the latency has to change massively was what Sergey, Marissa and Larry were obsessed with at Google. is how to get it faster, faster, faster, lower the milliseconds. People will search more if they made the search results faster.

And I think people will use AI more if you can get that, you know, interaction lower. I can probably get it by the time you have your test device. Yeah. It's going to be amazing. I mean, just the, I can tell you right now, the ability to order food faster. I do the same water for my family for the sushi rolls that we get, those like hand rolls. If I could just reorder that role, boom, and the faster I can do it, man, it would be amazing?

All right, listen, Jesse, you're awesome. When this thing launches, like, in three months, will you come on again? Of course, yeah. All right, here we go. And maybe we'll do it in person. Maybe we'll do some, oh, you know what? When you're ready to launch it, save a couple of demos, like maybe three and we'll do it live. Yeah, we'll just do live demos.

Why not? Yeah. No, no, but we'll do it live in person. I'll get a theater and we can just sit there and talk and you can show a couple of live things, demo or die style. All right, listen, if you want to learn more or if you want to, go work with Jesse and do crazy, awesome, rad stuff and cool shit like Steve Jobs did. How can people find out about your job openings?

Oh, simple. We have a career page. We're continuously updating in the roles. But just feel free to reach out to me via Twitter or feel free to send a contact at rabbit. com. Just go to our website, drop up the email. And the website is rabbit. com.

Yeah, rabbit. And make sure to check on the research page for the detailed breakdown of lamb, how it works. Yes, you have a really cool page about how lamb works. Yeah. So, yeah, just send us an email. We'll get it there. I mean, what are you hiring for? What positions are like the hardest to?

Yeah. So it's basically we need to expand most of the product related roles in every aspect because we're trying to catch up with the demand. And there's a lot of things that we're trying to. you know, built for the future. And more importantly, we want to set up a lamb research, like small land research team, just working for the future iterations of lamb. Are you in person now?

You're in person in Santa Monica, trying to get people in an office? Yeah, but, you know, if you're good, it can be flexible. Flexible, yeah. But you need to be really good. If you're really good, flexible, but if you're very good, let me tell you something. Living in Santa Monica, there's a lot of great restaurants. It's an incredible lifestyle, man. Just move and hang out with Jesse and Santa Monica.

It's one of the great places you live. My goal is to raise more money so that we can afford you guys. But yeah, that's my goal. You'll be fun. I think you'll be fun. All right, everybody. We'll see you next time in this week as startups. Bye-bye.