#29 - How AI could reshape human memory and attention - Bobak Tavangar

AI could change the way we remember, and the way we pay attention. 

In this episode of The Deep View: Conversations, Editor-in-Chief Jason Hiner sits down with Bobak Tavangar, CEO of Brilliant Labs, one of the most intriguing startups in AI hardware today. 

While trillion-dollar giants like Meta and Google race to define the future of AI glasses, Brilliant Labs is taking a radically different path: building in public, going open-source with both software and hardware, and centering their next product, the Halo glasses, around something deeply human. 

The focus? A conversational AI agent for your long-term memories and conversations. 

This isn’t just about smarter wearables. It’s about a bigger idea: 

+ Can AI help us be more present, not less? 
+ Could technology support memory, reflection, and intention instead of distraction? 
+ What does privacy look like when AI can recall your life? 

Jason and Bobak also explore: 

+ What he learned during his time at Apple 
+ Why AI hardware is one of the hardest frontiers in tech 
+ The challenging process of finding a co-founder 
+ Bobak’s philosophy on communicating on social media with purpose, not hype 

Bobak is one of the most thoughtful founders in the AI space, consistently elevating the conversation beyond features and into questions of values, agency, and human experience. 

If you care about where AI, wearables, memory, and attention intersect, this is a conversation you don’t want to miss. 

Subscribe to the podcast for more unique conversations with the brightest minds solving the biggest challenges in AI. 

And don't miss The Deep View daily newsletter. We don’t just cover AI — we decode it. In a world flooded with hype, we deliver sharp, no-nonsense insights that keep our audience ahead of the curve and help them put AI to work every day: https://subscribe.thedeepview.com/ 

Creators and Guests

Host

Jason Hiner

Editor-in-Chief of The Deep View

What is The Deep View: Conversations?

From frontier labs and enterprise platforms to emerging startups reshaping entire industries, The Deep View: Conversations podcast interviews the brightest minds and the most influential leaders in AI.

Jason Hiner (00:20.812)
Welcome to the DeepView Conversations. I'm Jason Hiner, and our guest today is Babak Tavangar, CEO of Brilliant Labs. Babak, welcome to the show. Thanks for being here.

bobak (00:31.086)
Thanks for having me, Jason. Great to be here.

Jason Hiner (00:34.048)
Exciting stuff. Well, Babak, your company Brilliant Labs is doing some very interesting things with smart glasses and AI. But before we get into that, I'd love to talk a little bit about you and your journey to get where you are. You're a startup founder. You've worked for Apple. You had a startup before you went to Apple. So how did you get into this? How did you get into your work founding a startup, how did you get into your work working on glasses and AI? Take us back how you got started in the tech industry in the beginning at all. And I would love to know just a little bit more about that. I know something about you, but would love to hear more about your journey to get here.

bobak (01:21.728)
Yeah, but there's a couple of different chapters. So, you when I was working my way through my undergraduate, I did an internship at Microsoft and it really got me thinking about those things that amplify human potential and how I wanted to spend my life and career sort of harnessing some of those forces. And the more that I thought about it, there were three that came to mind back then, undoubtedly there's more, but the three that came to mind were education, faith,

and technology. And that these three things had this tremendous power to unleash, to amplify someone's potential onto the world. you know, I didn't think I'd be very effective educator. You know, faith, I'm a Baha'i, so in the Baha'i faith, we don't actually have a priesthood, so that wasn't gonna be a viable career path for me. But technology was...

Fascinating technology was dynamic and back then this was sort of 2007 2008 it was beginning to touch all corners of the world. This was sort of around the know software is eating the world era and you know people were starting to see technology as this sort of layer that was going to touch all industries as much as it was a vertical industry itself and and so that was sort of what wrote my story.

at that early stage in technology and I never looked back. So the second phase of that really is, you know, working my way through my undergraduate and then graduate school and then starting my own startup, like you mentioned, and kind of realizing that there are some powerful software technologies. You know, back then we were working on graph search and graphs of course are

Jason Hiner (03:12.302)
Mmm.

bobak (03:14.206)
imminently useful in the AI era, especially for those of us working on memory, it helps to structure agentic memory. But we were also working with computer vision. Back then, we didn't have generative AI. So was a lot of machine learning, neural nets. And, you know, I was playing with these technologies and oftentimes within the confines of the smartphone. And it was the first time that I started to think, you know, for as

awesome as some of this software is, some of these technologies are so powerful, but we're beginning to bump into the walls of the form factor itself. That the hardware is really constraining what's possible with this software in our lives. And that was the first time, this was maybe 2016 or so, when I started to think about new form factors. And back then, Magic Leap, and there was another company called ODG.

There was a company called Meta, before Facebook rebranded as Meta. was a different group of technology, different group of companies, excuse me, and they're all working on AR. And back then it was sort of heavier headsets, console grade graphics in front of your eyes. And I started to ask around and try to figure out what some of those folks were up to. Fast forward again, a couple of years, I was at Apple and...

Jason Hiner (04:14.638)
Yeah.

bobak (04:35.852)
You know, there were rumors at that point that Apple was working on something in this space. And of course, now we know that that was the Vision Pro. But talking to colleagues internally at Apple, it was clear to me that it wasn't a solved problem there either. That, you know, these are challenging, hard problems that billions of dollars and some of the smartest people on the planet were still having a hard time solving. And so that gave me the confidence to start my own company. And so it was 2019 that I left.

Apple joined up with my co-founder Ben and we started Brilliant Labs.

Jason Hiner (05:11.118)
Very good. So what did you study in school and how did you end up at Microsoft for your internship? What did you do at Microsoft?

bobak (05:21.141)
Yeah, my undergraduate was in economics. My grad school was also in economics management, did a lot of research on the Chinese economy and how China was harnessing engineering, fundamental engineering and technology as part of industrial policy. So I'd long been interested in those sets of questions and technology sort of wove its way in and out. at Microsoft, I was a product manager on

the first wealth management piece of software that was being rolled out in China, Microsoft Money. I don't think it's around anymore, but it was going, maybe you remember it, was going toe to toe with some of the other wealth management software that was out there. Quick End, exactly. Exactly. So I was running product management on that as they were entering China through a lot of China's big banks.

Jason Hiner (05:58.626)
Mmm, yeah.

I do.

Jason Hiner (06:07.032)
quick in and other things like that, right? Yeah.

bobak (06:19.934)
They wanted to go through the banks to reach some of the newly wealthy in the country. So was fascinating and sort of a story as much of China as it was of software. So learned a lot. then when I was in graduate school, I had sort of a lightning bolt moment. I always knew since I was younger that it was inevitable I would start my own company. I come from a long line of entrepreneurs and I kind of got bit by the startup bug and

It just felt inevitable. It felt like the right time. And that's when I started my first company.

Jason Hiner (06:55.416)
So Baabak, how did, yeah, we also should say that you are in Asia. So today, your company's based in Singapore. You do a lot of work in China and Hong Kong. Where did you study and then how did you end up in Asia?

bobak (07:15.69)
So I was born and raised in the US. Born in Philadelphia, did my undergrad in DC and then grad school. I went across the pond as they say to the UK, so spent a good period of time in the UK. And then came to China when I was for the first time was during the Olympics. You know, that sort of brings us back. But 2008, that's right, it was it was a different era. There was a.

Jason Hiner (07:39.182)
2008? Yeah.

bobak (07:45.204)
a really exciting kind of spirit in the air. China was emerging and the world was coming to celebrate that at the Beijing Olympics. And that was my first foray into China. It was infectious, beautiful, fascinating. I, you know, just like I got bit by the tech bug and the startup bug, I got bit by the China bug too. you know, China and Chinese, it's one of those oceans. You can keep swimming for the rest of your life. You'll never reach the bottom. So.

Jason Hiner (08:10.894)
Mmm.

bobak (08:13.726)
That was the first time and I haven't looked back since.

Jason Hiner (08:18.018)
What was it attracted you to wanting to be there? You know, you mentioned what was happening, but did you do your internship in China or was it in Microsoft in the US?

bobak (08:29.226)
This was in Beijing. Yeah, so the internship was, that's right, the product group was local in Beijing. And, you know, I think rightly so. You know, they had to have some of the relationships on the ground, knowledge of the local market, and just proximity to key stakeholders to be able to localize the software appropriately and get that rolled out. So this was in Beijing. Yeah.

Jason Hiner (08:30.818)
Okay.

Jason Hiner (08:53.442)
Very good. Hang on one second, I'm going to pause for one second.

bobak (08:55.762)
Sure, sure.

Jason Hiner (08:58.166)
I just had to turn off a space heater. It's really cold here down in my podcast studio. Okay, back to it. So when you were at Microsoft, a lot of times, a lot of these big tech companies, they're really good at hiring their interns. Did you not had an opportunity to work for Microsoft and then chose to do your own startup? How did that work when you started kind of your first thing after college?

bobak (09:02.952)
Ha ha ha.

bobak (09:27.014)
Yeah, it was an interesting period. I was actually debating staying on longer at Microsoft. There was that opportunity. I decided to go to graduate school instead and there were a few reasons for that. I think by that point I felt that I had enough grounding obviously in where I was born and raised, which is the US. I had been exposed to China and.

learned the language by that point as well. My Chinese was fluent. And it just felt like the UK and Europe were gaps in my own lived experience, my own sort of tacit knowledge and sense of the world. And when there was an opportunity to go to grad school there, it was just felt like something that was right for me to do at that time. you know, big companies like Microsoft would always be there. And so, so I went and did my grad school there and then

straight back to Asia because especially back then that's really where the energy was and you know the the gravitational poles of the world were shifting and it just felt right for me to be back in Asia and just you know contributing to whatever was happening there.

Jason Hiner (10:31.384)
Sure.

Jason Hiner (10:42.158)
So you're in the US, spent a couple years in the UK doing your masters and then in Asia and you've been in Asia ever since. So did you, your first company after grad school that you started, was that also, that was also in Asia?

bobak (10:49.98)
That's right. Yeah.

bobak (10:59.068)
Yeah, so Infograph was based in Hong Kong and we were working on graph search and computer vision and sort of the intersection of those two things. So back then it was drone companies, satellite companies. We did some fun stuff with the mobile phones. We built this really neat core technology that was able to analyze what a camera saw and pin down a precise geolocation of something and then serve that up for a variety of use cases.

Jason Hiner (11:03.201)
Okay.

bobak (11:27.687)
And so it was really in that context and searching across that data, of course, as well. That's where the graph bit came in. It was in that context when we were playing with some AR stuff. And back then it was like really crude SDKs. It was a company called Wikitude and a few others, Metaio. I think they were bought by Apple. And, you know, so back then they were independent and developers like us could play with their SDKs. And I would hold out my phone and walk through central Hong Kong.

Jason Hiner (11:38.958)
Mmm.

bobak (11:57.207)
And we structured it so that we could see pinned locations of stuff visible through the phone's display because of what the camera and our core technology, geolocation technology was being able to pin down. And I felt for the first time both the magic of AR being able to elevate a reality that was not visible with my eyes. And at the same time, I felt the constraints of the form factor that

The hardware really wasn't suited, of course, to all day use. No one's gonna walk around with their phone going like this, and the battery certainly wouldn't even last that long. And so that was, yeah, that was that first moment when I started thinking, wow, there's something really special to this experience. And at the same time, it feels like it's just beyond our grasp. The hardware is not at a point yet where this is...

where this is real for humanity, but perhaps this is the next paradigm shift in computing. So that's what got the gears turning.

Jason Hiner (12:59.726)
So this is the mid 2010s, right? That's where we're at. And now just before that, 2013 was when Google Glass had arrived, right? And so it started to maybe tease this promise a little bit, but it also hit these constraints that you're talking about. you then aware that the fact that the hardware itself was still struggling to sort of keep up with the vision of what the technology could do?

bobak (13:02.481)
That's right.

bobak (13:28.795)
Yeah, actually, I Google Glass is no conversation about AR can happen without reflecting on the lessons learned from Glass. And I give Google a lot of credit, actually. They were pioneers in the space. They had the courage of throwing their hat in the ring first and really showing the world what sort of a slimmed down AR implement could look and feel like. I was.

walking through central London. I was in grad school at that point. I was walking through central London when I heard about this announcement Google was going to make and I ducked into a cafe and watched it on my computer and it was kind of magical. You I think like many AR enthusiasts back then watching them jump out of that plane and live stream simultaneously down to people, it was a different way of experiencing a computer in your life and a little bit mind-bending.

Jason Hiner (13:56.29)
Mmm.

Jason Hiner (14:13.166)
Yes. Yes.

bobak (14:22.542)
So that was definitely a formative moment. And at the same time, there were a lot of lessons learned. The form factor and the industrial design, a lot of the privacy concerns. So kind of on a sociological level, how people responded to that and the price point. I think there was a lot of lessons learned, but I definitely tip my hat to Google for having the courage of their convictions.

Jason Hiner (14:41.57)
Yeah.

Jason Hiner (14:52.77)
Yeah, so Bobak, for your first company, what was the product and how did the journey go with that and how long you did it? You did it for about five years, is that right?

bobak (15:03.343)
That's right. Yeah, yeah. So for Infograph, we were going for about five years. And then, you know, we ended up selling the technology to someone who had better use for it in the travel space. And I went to go work at Apple. But, you know, we started, like every startup, was a pivoting, winding journey. So we started really focused on graph search. And that was part of my interest in grad school, where looking at

complex interconnected networks of data, whether that was networks of people or weather systems, companies in the financial market. There's different ways to model this stuff. And one way is to look at the connections between them, the relationships and the density of those relationships as they evolve over time. you know, in especially in the computer science world, that's known as a knowledge graph.

And so this became just a fascination of mine, understanding connectedness, how all things are connected. And we started the company first to analyze that at the company level. And so we wanted to bring GraphSearch to private equity, hedge funds, people who were trying to understand how to quantify risk and value in a market. And our thesis at that point was, OK, well, you can do your

discounted cash flow analysis or DCF. And there's a few other ways that you can look at value and risk. But one way that people aren't understanding this is systemically. It's as a company is embedded within its ecosystem, what is the systemic risk to that company, but also the unrealized value based on its connections with those around it. you know, social networks like Facebook and MySpace and, you know, they had already been

executing on a pretty disciplined strategy of realizing this in the social media, in the social network world. But at the company level, especially in FinTech, this was still a very new concept. So we spent the first year or two trying to explore this. And eventually we decided to pivot. We brought it to news. So analyzing it, using natural language processing to break down articles about companies and about

Jason Hiner (17:02.254)
Yeah.

bobak (17:27.041)
other kinds of organizations and then structure those as a graph, as an interconnected web. And then we introduced computer vision and core location technologies to complement that because there were an emerging market of whether it was drones or satellites or autonomous vehicles or handheld cameras. On a phone, there was an emerging market of things that were making use of these data points. So we thought that we could cater to those.

companies and eventually we built a really neat little course suite of technologies and sold that off and I took my learning and went to Apple.

Jason Hiner (18:07.16)
So how did you end up, so you sold the company, you sold your first startup, and then you ended up at Apple. So you mentioned that you always knew you wanted, before you even started a startup, you knew you were gonna be a startup founder, like you had a pretty strong bent in that direction. How was it that you obviously interned for Microsoft and you said that you knew that you could always go to work for a big company. How did you end up?

With Apple and why, after doing your first startup, did you decide to go and work for a big tech company again for a bit?

bobak (18:43.278)
Yeah, I think at that point, it's a good question. You know, at that point, it felt like I had been running a, you know, I've been running the hamster wheel for five years. And, maybe hamster wheel is a bit, maybe it's not the right analogy or metaphor, but you know, the rat race, maybe that's a better way to put it. You know, being a founder is...

Jason Hiner (19:04.931)
Yeah

Jason Hiner (19:09.144)
Sure.

bobak (19:11.69)
It's intense, it's relentless and you know, the highs and the lows are similarly intense and oftentimes happening at the same time. And I'd been at it for five years and I think I was ready to learn something new and learn within the walls of a storied company that was building products on a really profound level of excellence and Apple.

very much was and is that company, especially in hardware. And at the same time, I think I was ready to not worry about where my paycheck was coming from the following month. And, you know, it's nice to work and just focus on, you know, what you're doing that day to advance the program without thinking about raising money for it or, you know, going out and persuading people to take a salary cut to come work for you on it.

So I think I'm working for a big tech company. It was as much a recharge as it was an opportunity to learn something. And this kind of brings me back to what we talking about earlier, where I was sufficiently fascinated with what the next paradigm shift in computing would be. And I had a hunch what it would be. And I knew that I needed to learn a lot more about supply chains, operations.

hardware design, hardware software integration. I just need to learn a lot more about those things. And there's few companies in the world, especially back then, that did it at a level of excellence that Apple does it. so, and you know, it was one of the best decisions I ever made in my life is going to work at Apple because the people that I met and the experiences, the amount of exposure to just different parts of the operation that I had, it was phenomenal. It was really excellent.

So it was exactly what I needed to equip me with the mindset to be able to start Brilliant Labs.

Jason Hiner (21:17.676)
Wow, so were you continuing to work, were you continuing to be in Asia when you worked at Apple and what did you work on while you were there?

bobak (21:27.35)
Yeah, so at Apple, was primarily in China, based out of the Shanghai office, spending a lot of time in 300,000 person factories, making iPhones across the country. So I was still Asia based. My Chinese language really came in handy as well. And at Apple, I was working on a software program within Apple operations. So...

Jason Hiner (21:45.39)
Mmm.

bobak (21:51.702)
That's where Tim Cook sort of, you know, he ran that part of the company. He was elevated from operations and they look across the entire supply chain. So it was a time when Apple was trying to think a lot more about, you know, what are some new ways that a company like Apple with its scale and leverage over the supply chain, what are some new ways they could think about gathering data from that supply chain, understanding the dynamics in real time across a few different metrics, whether it's people and labor.

whether it's data collected on the production line. Apple has one of the most complex supply chains in the world. So it was an opportunity, a data science opportunity really. So it was a fascinating insight, both into how products are made, but also how software and product design and other aspects of the company interface with that supply chain.

Jason Hiner (22:46.744)
So that work you did on interconnected ecosystems in your first startup clearly came in very handy. You put that to work when you went to Apple in the context of these very complex ecosystems that Apple had with Supply Chain.

bobak (23:03.486)
Absolutely. It gave me this framework for thinking actually, whereby I could jump into this new role at Apple amidst a ton of complexity. And it helped me to connect dots between stakeholders, whether within Apple or outside of Apple's walls in a way that helped me understand the complexity.

Jason Hiner (23:08.814)
Okay.

bobak (23:32.034)
It doesn't fit neatly into rows and columns on the spreadsheet or bullet points or however else one might typically seek to structure these things, but it really is a graph, a knowledge graph that can be analyzed as an interconnected whole. definitely, that framework for structuring complexity came in handy.

Jason Hiner (23:55.993)
So how long were you at Apple before you started thinking, you had this little bit of a break from startup life, before you started thinking about starting Brilliant Labs?

bobak (24:07.785)
I, you know, once a founder, always a founder to a degree. And I knew after my first company that it was a matter of time. So I was at Apple for close to two years, about a year and a half that I was there. So, you know, a brief stint relatively speaking. And I came into Apple already curious about, you know, this next paradigm shift, about new form factors.

And so I wanted to hit the ground running and use my time there to just meet as many people cross-functionally as I could, understand how these functions are knit together and just understand from a company that was, you know, in it is world-class at doing hardware, how it was thinking about new form factors. And so, you know, it was over my time there that

sort of this lightning bolt again, know, a lightning bolt hit me while I was sort of in this lecture hall there and alongside a bunch of my peers and hearing an instructor talk about the history of the iPod and how some of the learnings from the iPod, was a very necessary through line to the iPhone and that for iPhone to happen, there had to be iPod, something that was focused function and pocket sized and it generated learning among Apple and its

Jason Hiner (25:22.85)
Yeah. Yeah.

bobak (25:34.032)
suppliers that was necessary for iPhone to emerge. And there was something about that that just, you the lightning bolt hit and I had been trying to turn the thesis in my head, you know, what, what is that vector? What, what is the different path up the mountain for, for this new form factor? And that would drive what people do with it. That would drive how it would look. That would drive how it was costed because

I was looking at the broader industry working on it and it just, it wasn't quite happening. And despite the billions of dollars, despite the army of really smart qualified people, it just wasn't quite catching. And so I was trying to turn this thesis on my head and it was at that moment that I said, aha, it's gotta be something really simple and focused. It's gotta be, you know, the equivalent of what pocket sized would be for this new form factor.

Jason Hiner (26:19.022)
Hmm.

bobak (26:33.44)
and it had to sort of do one or two things really well and just start there. Had to be low power, it had to be extremely affordable and you know, sort of an iPod for AR and it was off the backs of that that you could begin to generate learning that would eventually lead you to something more complex and you know, full featured. So that led us, know, it led me eventually to start Brilliant Labs.

And that led us to do our first product, which was Monocle.

Jason Hiner (27:05.228)
Very good. So at this point, you also mentioned that you found a co-founder to start Brilliant Labs with you. Did you have a co-founder with your first company? And what can you say about like, what do you look for in finding a co-founder? That's a big question for a lot of people. A lot of people are out there today wanting to think about, should they embark on the...

on the founders journey and this question of finding the right co-founder is a big one when it comes to that. How has that been for you and how did you find that for Brilliant Labs?

bobak (27:41.203)
This is probably one of the most important questions. You know, my first company, I didn't have a co-founder. I worked with a team, but they were all folks that I recruited and hired and managed. It was something that, looking back, you know, when I was at Apple, I did a lot of reflection on the first stint and the first company.

Jason Hiner (27:49.441)
Okay?

bobak (28:08.944)
And it was one of the things that I wish I had done differently was find someone who, who complimented me in a way that co-founders ought to. And that's both a, like a skillset, you know, consideration, but also an emotional temperament type of consideration. And the reason being that in the really key decisions, you know, big and small, but especially in the really key decisions.

it's really helpful to have someone to consult with, to discuss problems with, especially someone who's been around since the beginning and has a sense of longevity in mind, and also a sense of the decisions that have come before and what we've learned from those that can be brought to bear on the one that we're considering right now. It's almost like a life partner in a marriage. And the ability to have that in a startup is really critical.

Jason Hiner (29:00.418)
Yeah, yeah.

bobak (29:07.25)
Being a solo founder, I was just in my own head all the time, trying my best to make good decisions and learn from what came before, but without another person to have the clash of opinions with, to seek truth with, to reflect on what came before with, and just to bring that different perspective. Each human being brings a different perspective, and there's something really important about that. And so...

for Brilliant Labs, I knew that I needed to have that. I couldn't do the solo founder thing again. There was just too many gaps in my own experience and judgment. And I needed to have that complimentary skill set and perspective. And that's very much my co-founder, Ben. So Ben is an industrial designer and his temperament is so different from mine and it just works beautifully.

for all of my staccatos, he helps to kind of ground me. And we have a really beautiful, you know, sort of dialogue between us when it comes to key decisions and reflecting on what came before. And frankly, sometimes it helps to just laugh too. You know, we'll go to dinner sometimes after a long day at the factory and we'll just think about what came before the...

you know, the face plants and the mistakes, but also just the funny things you see when you're making hardware. And we just chuckle about that stuff, you know? It's almost, it's like an old married couple that can just chuckle about what came in the decades before. And there's something therapeutic about that, and then there's just something really important. That support and the complementarity. It just rounds, you know, the team off in a way that helps the...

Jason Hiner (30:40.44)
Yeah

bobak (30:55.826)
You know, the stool stand firm on a few different legs. you know, in terms of, you know, whether there should be a co-founder in place, absolutely. You know, that was my learning. In terms of, you know, who that co-founder should be, you know, for me, was very serendipitous. I'm very fortunate. It was colleagues of mine at Apple who knew Ben and were able to introduce me to him. They knew me well enough. They were excited that I was...

imminently leaving and starting this company and you know, they basically said, look, you want this company to be, you know, I was telling them sort of the rough outlines of what this new company would do. you know, coming from Apple, I think there was an appreciation for a company being design led. so these colleagues of mine said, look, you want to be collaborating with a really top notch industrial designer from day one. And we know just the guy.

Jason Hiner (31:43.822)
Hmm.

bobak (31:54.533)
And so they introduced me to Ben and we hit it off. mean, there was a nostalgia factor. sort of brought me back. He's from the UK. So he brought me back to my days living in the UK and just how much I enjoyed talking about products and ideas with friends back in the UK. And we literally started with sketches on the back of napkins in a cafe in central Shanghai.

Jason Hiner (32:07.246)
Hmm.

bobak (32:21.655)
and just dreamed of different paths up this mountain of wearable computing, different from what had come before and what might that be. It was a very exciting time. And through that process, we got a sense of each other, creatively how we were different, but also how the jazz came together, how we were able to riff on each other's ideas, and also how thoughtful we were able to be together, which is...

I've learned it's just very important. How well do we listen to each other? Do we tolerate moments of quiet? Do we tolerate tentative ideas that are maybe half-baked but no less important? So a lot of really fine points of collaboration like that. We just got a sense of each other over months of just meeting multiple times a week and sketching and getting some 3D printed and...

testing stuff on our friends' heads and glasses. There was a lot of ethnographic work. We would go on the subway and we would just observe how people use their AirPods or their phones. We would observe that people would want to take out their phones and open the camera app, for example. Sometimes it wasn't to take a photo. Sometimes it was just to pinch and zoom so they could see something with higher fidelity.

Jason Hiner (33:46.414)
Mmm.

bobak (33:46.599)
So we would try to observe little corner cases, unintended corner cases that people were trying to use the current batch of technology for. And it was speaking to what new form factors might do for people in a way that was more purpose built. So a lot of that stuff over those early months before we had enough conviction to start the company together, formally set up the entity and we were off to the races.

Jason Hiner (34:15.64)
So you're now about to launch your third product, but your first product was Monocle. And my sense is this happened pretty quickly, especially, know, founders sometimes they embark on something. It may take a couple of years before anything comes to market. At the least, you know, you're often working on things for a pretty long, you know, runway before something hits market. My sense is you all came to product, brought a product to market.

pretty rapidly and I had the sense of like, you you wanted to put something out there to sort of learn as quickly as you can. Is that correct? How did you approach, you know, releasing this first product that was very kind of prototypical, almost feels like a lab sort of experiment that you did with your audience or with your customers?

bobak (35:06.916)
That's, yeah, no, I think that's a great way to frame it. you know, we're open source, but part of the ethos of the company is to be open source. And so part of that ethos is to be very experimental together with our community, knitting new technologies together, new ways of thinking about a problem and

putting that out there and it's as much to see what people build with it as it is to learn from the new opportunities that we didn't foresee and where we should evolve the platform. we actually, it's interesting that now with hindsight, it kind of looks like it was quick because we actually felt it was very slow. We started the company in 2019 and then COVID hit and...

It was, I often joke with people, it the best timing to start a hardware company. On the one hand, were thankful at least that we were not in the middle of mass production and then everything shut down and so all of a sudden we couldn't ship units anymore to people who paid for them. So we were thankful, but development was labored. It was tough because there was a few years there where we couldn't travel.

to be in factories or get stuff made quick enough. If factories were open, they were on sort of a staggered schedule and lead times were really drawn out. And then around 2021, 2022, there was this really terrible semiconductor crisis. for folks who weren't in hardware, maybe it sounds like, maybe it doesn't sound familiar, but...

you know, the, the, auto industry, think, I mean, this is, there were people kind of, you know, pulling their hair out, trying to wonder the origins of this thing, because it felt like a plague of locusts. But the auto industry, had, had suddenly sort of, sales had collapsed in the beginning of COVID. And then as, as lockdowns were lifted and as, you know, demand, much to people's surprise, demand just shot back up.

Jason Hiner (37:03.266)
Yeah. Yes.

bobak (37:23.467)
And that included, you know, materials to build homes, included consumer electronics, it included automobiles. So cars, it was hard to find cars, even secondhand cars. And I'm sure you, that's right.

Jason Hiner (37:37.816)
That's right. They couldn't finish them. I can tell you a funny story of like not far from where I live. There was a whole football stadium that was full of F-150 trucks that they parked all across the stadium because they couldn't get the semiconductors, which is the last part that they needed, you know, the chips to finish like two, two aspects of this truck. Like the trucks were all built, but they couldn't roll them out because they were still missing these two parts that actually made them work. And they would not work, you know, without those last two.

bobak (38:06.785)
Wow. Wow. That's interesting. Actually, I hadn't heard that before, it's from the other side of it. You know, we were a small startup trying to buy memory chips and processors and FPGAs and different sensors. And the word on the street was that a lot of, know, there were big companies like Apple and Microsoft, of course, just hoovering up supply. And they were often at the top of the list. They were the front of the line if they needed something.

vendors would favor them. But it was auto companies who were short of supply for critical chips and they would risk buy. So they would over buy and just, you know, the market, there was just so many gravitational forces pulling supply away from startups like us. And, you know, that meant that we would, if we could get anything, we would every single day, I was up till 3 a.m.

Jason Hiner (38:45.015)
Yeah.

bobak (39:04.597)
monitoring multiple secondhand distributor websites, just trying to find a dozen of something. And, you know, it was chaos. so all that's to say, when we started the company and we started working on Monocle, there were, you know, aside from COVID and aside from the semiconductor crisis, and there were a couple of other things that happened as well that just really, it...

It elongated the development cycle of Monocle. We finally shipped in, you we announced it in late 2022 and we started shipping at the very beginning of 2023. You know, that very well could have happened a year, year and a half earlier. But it proved us, you know, for nothing else to ourselves that we could survive through the plague of Locusts.

Jason Hiner (39:51.854)
Okay.

bobak (40:02.521)
We could learn, we could be resilient. We could persevere through these different crises that set the company and really sort of focus on the things that were within our control and try to execute discipline against those and just roll with a variety of punches until we get a product out and just stay focused on the goal of open source AI focused computing and just learning together with our nation community.

Jason Hiner (40:03.01)
Yeah.

Jason Hiner (40:31.49)
Wow, how about the funding for the company? you launch the company and you survived a couple years, I'm assuming you must have a funding journey that goes with that story as well that enables you to sort of have that runway until you could get your first product to market.

bobak (40:47.827)
Yeah, I mean, in the very beginning, was self-funded. and I just, you know, our costs were low. It just the two of us. And, you know, eventually started iterating with a development partner in China. And so we just self-funded a lot of this stuff. And then, you know, we wanted to hire a few people on the team. We wanted to expand operations. So there was a little more material procurement. And so our costs...

were a little higher and so we needed to raise first capital. So we went to a few early stage VCs, a couple of angel investors. I've never felt comfortable with the friends and family thing. And so I've always felt that there is a whole class of investors who bring value to the company because they're seasoned investors, but also they've got the risk capital to be able to put into startups. so it's better to...

Jason Hiner (41:37.219)
Yeah.

bobak (41:43.99)
sort of look there rather than friends and family. So I've just always been of that mindset. Until we raised, you know, a few.

Jason Hiner (41:49.602)
you fund your first startup? Sorry, Bob, how about how does that compare to the way you did your first startup as well?

bobak (41:52.362)
Yeah, yeah.

Very much the same, very much the same. know, avoided the friends and family route and just went straight to angel investors and, you know, one or two early stage VCs. And, you know, it starts with a couple hundred thousand dollars, you know, on a safe note. And then, you know, it jumps maybe to a couple million dollars and you sort of go through the different stages of funding. But in our case, when we started Brilliant, it was fumes for years.

It was, mean, through COVID, there were long stretches where, you know, we just, couldn't pay ourselves. And we had to, we had to make every dollar stretch and, you know, negotiate really hard with vendors. And, you know, we, we, that was another aspect of learning how to survive was not just on the development front with making something, but it was also learning how to get creative to just make dollars stretch to continue paying for it.

So, you know, part of that was COVID, but really part of that, it was, you know, we were doing consumer technology and as we learned really quick, most investors, they're not seeking to invest in hardware, much less consumer hardware. And that's because software has these wonderful economics whereby you spend a bit of time, you can build something, you can push it out.

at near zero marginal cost and there's like mature distribution at this point, whether it's something web-based, whether it's something that goes through an app store, the distribution is just immediate and it's to billions of devices, billions of people if there's the demand for it. And you can iterate on that. You could pull all-nighters and push pixels and squash bugs and just continue pushing updates. And it's again, near zero marginal cost.

bobak (43:52.128)
And there's, know, mature cloud hosting to be able to scale up and down. So your costs sort of scale with your adoption, with usage. And, you know, as long as you can get people to pay you for that, then you're off to the races. So a lot of investors we discovered, they love the economics of that. Hardware, not so much. Hardware, there's few companies that, you know, that do that well. They tend to be very large, extremely well funded. You know, think of Apple.

Jason Hiner (44:11.586)
Yeah

bobak (44:21.479)
And there can be good margin on hardware, but that's not in most cases. We learned that actually most of the time, most hardware devices, game consoles are great example, they're subsidized at a loss and you try to make up money on the long tail of games that you put out and sometimes it works, sometimes it doesn't. Apple's one of the few that makes money off of phone hardware. think famously in the phone handset market, they maybe have

10 to 15 % of market share, but they've got 80 % or 85 % of the profits. you know, it's again, it goes back to Apple and their model and there's only one Apple and everyone else is just trying to figure it out. So how much less, you know, a startup, right? Like without that leverage over the supply chain, without, you know, the fit and finish of an Apple product. And so I think investors and you know,

Jason Hiner (44:56.494)
us.

Jason Hiner (45:09.378)
Yeah.

bobak (45:18.334)
You could argue, right? We saw they look at that. And I think many folks say, you know, there aren't too many apples or Teslas out there. Why this startup? You know, and especially if it's if it's consumer hardware, consumers are very fickle. It's you know, it's hard to do million unit volume, especially when the market is so unproven as in our case, you know, our market is very nation. It's like.

very early innings, not even the end of the first inning for our market, and it's still very much becoming. And that's risky in and of itself. So there were a few reasons, I think, why, you know, it was tough to raise funding early on, but we just needed to roll with that punch. We just needed to learn, hey, keep our costs low, iterate quickly on a dime, and learn how to make forward progress on fumes.

Jason Hiner (45:53.976)
Yeah.

bobak (46:16.694)
And we did that.

Jason Hiner (46:19.63)
So talk about what Monocle is. It is what it sounds like. It's a one eye, essentially, AR device that did some interesting things. So talk about what that product was and what you learned from it.

bobak (46:35.222)
Yeah, so exactly. Monocle is a little clip-on device. It had a camera, microphones, capacitive touch sensor. Monocle was $299. Yeah, $299. And it was kind of cute. It looked like a little one-eyed Android. fit in your pocket. And back then, we didn't know a whole lot about making glasses. And the production processes...

Jason Hiner (46:43.278)
How much did it cost? forgot. I forgot how much was it. Yeah. OK.

bobak (47:04.77)
the design, the ergonomics of the face, there was just a lot of new learning for us and we recognized that. And so we wanted to dip our toe in the water and we also wanted to plant a flag in the ground. We wanted to show the industry that you can think differently about this form factor. And more importantly, it ought to be AI focused. We were the only company back in 2019 when we started, we were the only company, not only that was doing it open source, but that made this

back then what seemed like a kind of crazy claim, we literally got laughed at when we would say, these new form factors ought to be AI focused. They ought to be about doing AI inference over what you see and hear. And people would laugh. People would say, we go to these conferences and people would say, you know, did you guys like, did you miss something? Because these form factors, they're supposed to be about.

seeing a big whale flying through the sky or a T-Rex dancing on the table in front of you, know, something fantastical, console grade graphics. That's what this is about. It's not about AI. People would literally tell us, what does AI have to do with glasses? What do these two things have to do with each other? You guys miss something. So that was the early days of the company. You know, we would be on investor calls. We would be talking to factories. We were trying to persuade people to work with us and we would get that reaction. Wait, AI and

Jason Hiner (48:13.378)
Mm.

bobak (48:28.532)
glasses? Like, what do those two things have to do with each other, guys? I don't get it. And you fast forward today, and now it's sort of the obvious thing in the room, right? It's all about that's right. That's right. So there was a lot of that, that we had to sort of work through and just hold to our own convictions. But the market came around and GPT published the API at the end of 2022.

Jason Hiner (48:32.043)
Wow. Wow.

Jason Hiner (48:38.444)
It's all about AI. The glasses are, yeah.

bobak (48:57.394)
at the exact time that we announced Monocle. And so, you know, we knew that this was gonna be big. We had this AI thesis back then, it was sort of computer vision and machine learning. And then the GPT API was published and we knew, okay, the game is about to be upped. Our thesis is about to, you know, start being validated on a whole new level at a whole new speed. And so we got to be on this.

And so the team very quickly did a GPT integration with Monocle. And that meant that you could be wearing it clipped onto your glasses and be asking a question as you're walking around the world hands-free and see GPT responses printed in front of your eyes because it had a little color OLED display. So.

Jason Hiner (49:38.254)
Wow. So OpenAI, released the GPT2, I think, at that point, right? GPT2 API before they released ChatGPT, which, of course, took the world by storm in November 2022. Yeah.

bobak (49:50.334)
That's right. That's right. Yeah. And so we...

Jason Hiner (49:54.727)
Good timing, like that's a key to startups. Like timing is always key and you had the benefit of some amazing timing at that point.

bobak (50:02.292)
It really felt like it. felt like we needed to survive in the wilderness during, you know, 2019, 2020, 2021, most of 2022. And that if we could do that, there was something good for us on the other side, you know? And so at the end of 2022, we finally made that announcement. And then we got this gift from OpenAI. And, you know, as we all know, the industry just went nuts. So that very much, you know, characterized

know, Monocle and its journey. And all of sudden the discourse was all about AI and because we were open source, people were suddenly thinking about, well, you know, if these guys could do an integration between the GPT API and that piece of hardware, what else could we do here? And so then this notion of AI hardware started entering the lexicon. And, you know, we had a nice little piece in Bloomberg and

And then we announced our second product frame, which was our first pair of glasses, color display, camera, know, many of the same sensors as we had on Monocle, but it was in a really thin, lightweight pair of glasses. And by that point, AI was multimodal. So GPT-4 came out with vision and the glasses could see and not only hear. And, you know, it started to become obvious to people what the thesis should be.

Jason Hiner (51:28.334)
So was 2024. So Monocle came out, had this really great timing with, you know, chat GPT and then GPT2. And so 2023, you really spent learning about how that could work. And then your next product.

The frame which we also covered in 2024. We covered when I was at zD net we covered it and I remember Carrie one interviewed you for for that piece and We were we were quite impressed. I remember that you know, this was just one things like, know The first glasses that would come out from some of the big players, you know We're not very good to be honest the ones from Amazon and Metta and us and not just us when I was at zD net, know We reviewed them. were like these are

these are pretty terrible, we can't recommend anybody buy them. And the fact was it was very clear that AI was bolted on and was not sort of core to the product. But by that time the frame came out, you all were already sort of on your second generation of this. So talk about what the reception was for the frame and what you learned from doing that product. And at that point, is open source hardware starting to sort of be received?

bobak (52:36.05)
That phase, you know, when we announced frame and all of our learning thereafter, that was a really interesting phase for us. You know, in many ways it was, it was like, you we, we felt like we'd caught a tiger by the tail and

When we announced it, it was actually right after the reviews of the Apple Vision Pro came out. And I think they were less than stellar. And then we announced Frame. I think, you know, it just sort of hit this, it resonated at scene because you had a much larger, heavier, expensive product like the Vision Pro. And then you had an order of magnitude more affordable and thinner and lighter device like Frame.

Jason Hiner (53:07.903)
Mm-hmm.

bobak (53:32.506)
announced within sort of a week after and I think that just, it blew up. We thought we would only do sort of, you know, a handful, maybe a few baker's dozens of frames that year. Again, all in the service of just, let's just learn, you know, this isn't even a market yet, especially back then. You know, we would just tell ourselves, this isn't even a market yet. We have so much to learn about what the use cases are gonna be and...

you know, what this community dynamic is gonna look like. Let's just do a few hundred of these and not focus on sort of juicing the lemon and maximizing what this can be, but like just learning. And all of a sudden it blew up. We announced it and there was a bit of a reality around it. We got a lot more press coverage we thought we were gonna get. And I remember it was during Chinese New Year. We should have thought about the timing, but it just went nuts. Our sales went, you know, vertical.

And I was calling all of our suppliers in China, trying to describe to them in Chinese, trying to tell them, look, something's just happened. And I know that you're on New Year break and you're with your family and I don't want to interrupt you, but I need your help. And so, we were very fortunate. had a lovely group that we were working with on frame across China. so, we really tried our best to ramp.

Jason Hiner (54:38.83)
You

bobak (54:58.009)
supply as fast as we could, we were stretched in terms of the supply chain. We realized quickly that there were a few key bottlenecks, know, the geometric prism that the OLED panel was bonded to, which was placed inside of the right front frame in front of your eye, that that was heavily supply constrained. And so we needed to...

Jason Hiner (55:19.854)
Yeah.

bobak (55:26.929)
That was a whole endeavor to just try to figure out how can we add more people to that process? How can we cut unnecessary steps? How can we get more supply coming in to try to feed into that process, material, et cetera? And our ability to deliver against pre-orders elongated much, much longer than we thought it was gonna be and customers got frustrated and...

you know, that was all on the software, the hardware and production side. On the software side, we also quickly hit the reality of, my God, generative AI is exciting, but wow, is it subject to hallucinations and hiccups and, you know, models would hear a moment of silence and they would respond by saying, for, you know, thanks for watching, like and subscribe. That's what models, because most of them.

Jason Hiner (56:08.354)
Yeah.

Jason Hiner (56:17.92)
man.

bobak (56:20.027)
They were trained on YouTube videos and said they would hallucinate. Thanks for watching, like and subscribe. And customers were getting these responses and like they would send us screenshots. Guys, what is going on here? You know, why is it responding to me this way? So there were all kinds of things that we learned. But it, you know, it rocketed us to this new echelon that we never thought we would be playing at.

Jason Hiner (56:33.602)
Yeah.

bobak (56:47.297)
And at the same time, there were a bunch of other companies that were in the same AI hardware category like Humane, like Rabbit. There were a few companies, we'd been in touch with them because we were sort of in the same cohort that year. they had a really hard go of it that year. so there was so much learning. And that's exactly what me and the team needed. We needed this sort of trial by fire that stretched us. And that's what we got with Frame.

Jason Hiner (57:04.974)
Yeah.

Jason Hiner (57:17.422)
So what size is the team by that time?

Can we talk a little bit about the design of the frame? Because it had a very specific design that looked different than the other glasses. So it had very round frames that were quite interesting. What was the size of the team that it took to pull off that product? And when you got a bunch of orders, did you have to hire a bunch of people to help meet the demand? That's 2024. What did all of that look like in terms of how you decided on the design and then what made the team to build it?

bobak (57:51.93)
Yeah, on the design side, we wanted to go for something that looked different, that for posterity's sake would stand out as differentiated, dare I say iconic. So we looked to the past. We looked at Winston Churchill, Mahatma Gandhi, John Lennon, people who had these crafted round frames who...

Jason Hiner (58:18.744)
Yeah.

bobak (58:20.601)
were cultural icons, cultural, political, people who had shaped our world. And we felt like we wanted Frame to sort of try to bring a little bit of that cultural resonance to the conversation because this was the first pair of AI glasses. There were no glasses that, certainly in the marketing, no one called them AI glasses yet. And so we were the first to kind of come out of the gate and say that now everyone's AI glasses are everywhere.

Jason Hiner (58:42.862)
Yeah.

Jason Hiner (58:48.654)
They're all AI. They're not smart glasses or AR glasses anymore. They're AI glasses, right? In most cases, all of the competitors. Yeah. Yeah.

bobak (58:53.081)
That's right. That's right. Exactly. so, you know, we felt that again, just like Monocle, this was its own flag in the ground. And we sort of wanted the industrial design to try to speak to that a little bit. On the staffing side, yeah. I mean, in addition to calling up all of our suppliers during Chinese New Year, we knew that we needed more support staff because suddenly, you know, we have a pretty vibrant Discord server and

It's mostly developers, people who come for the open source aspect and want to just hack and build cool stuff. But, you know, we just knew that that rowdy community, as exciting as it is, we needed some reinforcements so that as the core of us were just really focused on trying to get the product out the door, there were folks who were staying in touch with, you know, concerns the community had, questions that they had.

And so we needed the beef up there. We needed to add a few people on the software team to parallelize a few tasks. And then we needed one or two reinforcements on the hardware team as well. There were a few things, especially FPGA development, which is a very specific kind of development process. We needed someone who was bringing a bit more expertise there to knit a few things together before we push the product out. So yeah, it stretched us within a very short period of time. We just needed

Reinforcements across the board, materials, suppliers, and personnel.

Jason Hiner (01:00:27.086)
How big is the team at that point?

bobak (01:00:29.325)
Yeah, at that point, the core team was about four people. And then we had an extended group of folks beyond the core four that were maybe eight, eight or nine people. Yeah.

Jason Hiner (01:00:42.616)
Okay. Okay. So that was frame. then now, you know, you've announced the next product that's coming, which is Halo and Halo will be released in 2026.

What was the journey like from the frame to Halo? Actually, before I want to ask another question about the frame, because when I think about the frame, there's also a design choice that I think about. I think about it had this very bright orange charger that almost looked like a clown nose and you like, it sort of...

bobak (01:01:14.114)
Yeah. That's right. Like Mr. Potato Head. Yeah.

Jason Hiner (01:01:18.606)
Yeah, yeah, like Mr. Potato Head knows and it like sat on top of those and a lot of people sort of in the reviews, I remember people sort of chuckling at that, maybe even mocking it, loving it, laughing at it, but you couldn't ignore it. Like it was so unique and interesting that it and when I think of the frame today, like going back to 2024, that's still what I think about.

bobak (01:01:28.395)
Yeah. Yeah.

Jason Hiner (01:01:42.126)
ultimately is that very unique charger. Tell us a little bit about that. Was that a conscious choice? Did it work out the way you hoped? Yeah, what did you learn from that very interesting idea?

bobak (01:01:55.303)
Absolutely. So Mr. Power, which is what we called the charger, which would sort of referenced Mr. Potato Head. It was designed to be whimsical, to bring a smile to people's face. You know, we, and this is still something that we think about a lot. As consequential as these technologies are, especially artificial intelligence, we also didn't want

Jason Hiner (01:02:04.565)
Okay.

bobak (01:02:23.863)
people to be so self-serious about this stuff because part of learning about these new technologies and exploring these technologies requires a sense of play. It requires a sense of whimsy and sort of exploratory mindset. so we always, know, especially when we were doing frame, but also now and in the future in both our hardware and our software, we're always thinking about how we can do little things.

like Mr. Power is a great example, how we can do little things to just bring a smile to someone's face and just generate that sense of whimsy and play, that notion that nothing is so precious that you can't peek around the corner or you can't sort of chuckle at it, that these technologies should be there for our exploration and play.

Jason Hiner (01:02:54.668)
Yeah.

bobak (01:03:21.687)
So that was, yeah, very much like, you we wanted it to be iconic and something that people would chuckle about for sure.

Jason Hiner (01:03:28.632)
Very good. All right. So you talk about you've got your next product, you know, sort of loading to come out. Halo in 2026. Talk about the journey from the frame to Halo. What did you learn from the frame? You see the industry now sort of moving in the direction of where you all had pointed the company earlier, sort of moving in your direction. But also you've got now these giants, Meta, now Google, you know, others.

bobak (01:03:36.033)
That's right.

Jason Hiner (01:03:56.974)
really coming in the direction that you know we expect Apple to also sort of move in the direction what did what does that mean for for what's next for you all and then how did you decide on what the new iteration of the product would be and I believe you're also wearing halo you know as well at the moment so tell us about that

bobak (01:04:15.268)
That's right. Yeah, so this is Halo. It's a lovely lightweight pair of glasses. It's about 40 grams. So I can wear it all day. It has 14-hour battery life, so it lasts as long as I can wear it, which is as it should be. Similar to Monocle and Frame, there's microphones. This time we have two microphones. There is a color display.

which is distinguished from many of the other glasses out there which have a monochrome green. So we just feel like, you know, from the day you're born, you live life in color and that's how your product should show up for you as well. So it's a color OLED. Low power camera, an IMU, which sort of is the sensor that picks up your directionality and your acceleration, so your movement. And on Halo, we've added speakers. So there's two bone conduction speakers, one on either side.

and that pipes spoken assistant responses or sound effects, whatever you want, so you can also hear things. So taken together, it is a pretty neat little device and we've packed a lot into a really tight lightweight form factor. And it's designed in service of what we think is going to be a really important use case, which is all-day AI inference, but especially for memory. So...

Memory is this really important, existentially important facet of what it is to be a human. And our memory is nonetheless fallible. And I think the older you become, the more you realize that. And we just felt like it was meaningful as long as we could do it in an extremely privacy-oriented, like a privacy-first way. We think that's critical. As long as we could do that, we felt it was a very...

Jason Hiner (01:05:43.949)
Yeah.

bobak (01:06:07.205)
a novel thing to be able to do AI memory that complements the memory that you have of your own life. So that means that in a way that you can trust, an agent is able to do inference over images and audio that's captured over the day and index that in a way which is, you know, even if the data was leaked is not something that another human being could see or listen to and make any intelligible sense of.

that it's just, it's saved as a string of numbers, all the inference is happening, you know, before it gets to, it gets saved on the server and then the original source media is tossed away. We actually don't save any of that stuff. We don't want to. So, you know, that hopefully gives people the sense that, you know, you can trust this agent as it's learning and indexing it for memory, but you can feel a sense of security as well that, you know, the data is not in a form where

Jason Hiner (01:06:47.394)
Mmm.

bobak (01:07:03.701)
you know, the memories of your life could be read by someone else, which is a very scary dystopian idea. And so, you know, we just felt that glasses could be and should be oriented around this experience. So we did Halo really in service of that. And of course, kept it all open source as well. So for us, I mean, that's not only software, that's also the firmware and the hardware. you know, electronics,

Jason Hiner (01:07:28.462)
Okay.

bobak (01:07:30.955)
mechanical, optical, all the design files for the hardware, the bill of materials, the different components and things, that's all open source. It all sits up in GitHub because we just don't feel like in 2025, 2026, we just don't feel like it's a competitive advantage to hold this stuff close to your chest. We think that actually there's more of a competitive advantage in creating a community around what you do.

A great way to do that is by sharing the knowledge that you've been able to develop through the process of building it and then inspiring others to build on top. So then it becomes this sort of mutual learning process. And it's just so much more fulfilling, we find, than just making something closed and trying to make a buck off of that. It just feels like bit of an outworn idea.

Jason Hiner (01:08:24.568)
So disseminating knowledge, it's interesting, you're the first founder I've talked to that sees like disseminating knowledge about what they're creating to create community around it. Seeing that as a competitive advantage is a really unique novel concept.

bobak (01:08:43.764)
Yeah, I think...

Jason Hiner (01:08:44.48)
not around, sorry, I should say to you, Baba, not for open source itself, but for a hardware company, you know, building a product and trying to, you know, build a commercial, you know, a business off of it as well. Obviously, the concept itself is part of open source and that's an inspiration, was an inspiration for what you're doing.

bobak (01:09:01.748)
Yep. Yep. It's absolutely the case that we just feel that we're living in a slightly different age now where, you know, empowerment and knowledge dissemination, that these things engender not only good feelings about a brand, but they also create community. They create an ecosystem and it is more powerful for people to be

attracted to that ecosystem intrinsically because of the value that's generated there rather than locked into that ecosystem because they're engaged in a bear hug. so, and so, know, we, and this was part of, you know, my learning from Apple as well was, you know, Apple does a phenomenal, they do phenomenal work in so many areas. And at the same time,

Jason Hiner (01:09:43.424)
yeah.

bobak (01:09:57.758)
you know, and we've seen this in the news the last couple of years, it just feels that there is an opportunity to still strive for that high standard of execution, but do so together in community. You know, bring people along with you in the journey, the knowledge you generate, the things you learn, share that with others. We just think there's so much more, you know, that can be a multiplier of value.

Let alone it's again great for the brand. You know, this is not at odds with doing good business and we just think that we're living in this day and age where this is possible now.

Jason Hiner (01:10:32.078)
I want to double click on that a little bit more, but before we do, for people that are listening to this that are interested in the product, how much will it cost? When is it coming out? When will they get to know a little bit more? Where do they find out more?

bobak (01:10:48.137)
So it's early Jan and so we're in the midst of ramping production on Halo in southern China. So the product will start shipping in a few weeks, but later this month. so Halo today costs $349, which is not too dissimilar from the $299 of Monocle and Frame was also priced at $349. And it's...

super lightweight and could very well be your next pair of glasses. So check it out.

Jason Hiner (01:11:23.758)
So you can get them with prescriptions or without both? Yeah.

bobak (01:11:27.356)
That's right. Yeah, we so and that was actually something that we wanted to learn more about this time around as well. You know, each product we've identified things that we needed to learn. And so with frame, we weren't able to really execute on the prescription side of things the way that we wanted to. So with Halo, one of the things that drove the design the way it is, is, you know, the OLED, the display is internalized in the top of the front frame so that it's easy to just.

pop a prescription lens in and out or a sunglasses lens or a transition lens, whatever you want. And so we've partnered with an eyewear company that does the full range of any flavor of prescription needs, sunglasses needs. And so when people purchase the device, when they purchase Halo on our website, we then point you over to them at a dedicated site that they've set up to order whatever prescription you need and they ship it to you, direct to your front door.

Jason Hiner (01:12:23.756)
good. the the glasses themselves are the frames are now square. So you moved away from this circular form factor. Very unique that you did with with the frame to halo is now square. It is only one style. Right. And maybe talk a little bit about that design decision because clearly you did it very purposefully in the way that you you made the design with your previous product.

bobak (01:12:48.816)
It's okay if you face it.

bobak (01:12:54.382)
Yeah, I mean, this was sort of, again, in service of the placement of the display. Sitting up top here at the front frame, it meant that the top of the frame needed to be plateaued somewhat to align with the rectangular shape of the display. so plateauing up top, we considered

Jason Hiner (01:13:10.583)
Okay.

bobak (01:13:21.0)
you know, taking the rounded frame design and just sort of plateauing the top of that. But, you know, I think we felt that we wanted to evolve and play with a different kind of form factor a bit. And so we went with something a little more of a traditional squared off design. And, you know, who knows what we do next? Maybe it'll further evolve from there.

Jason Hiner (01:13:41.902)
Okay, okay. So in service of the technology, the frame is in service of what needed to be with the technology. you also went from, correct me if I'm wrong, but the frame was a monochrome green display, and then this display is full color.

bobak (01:14:01.128)
No, actually, a frame was full color and so was monocle. So that's something that we've always felt very strongly about is that, you know, these products should always be full color because, you know, again, life has lived in color and we just think that the product should show up for you in the same way. So, yeah, we've always been about doing full color in front of the eye.

Jason Hiner (01:14:08.514)
Okay.

Jason Hiner (01:14:23.274)
Okay, forgive me, I see so many of these glasses and there was the first display, there was a bunch of other products that had monochrome green displays that came to market. so, no, that's very interesting that it's always been full color and it's impressive keeping the product at 349 and to have a full color display. Like we've seen the Meta Ray-Ban display

bobak (01:14:27.463)
Yeah.

Jason Hiner (01:14:51.63)
you know, go with a full color display, but now the product is $800. How can you keep it at such a low cost? And you've also maintained that sort of, you know, 200, sorry, 300 to $350 cost point. I assume that that's for a reason. You see that as the sweet spot. How are you able to keep it there?

bobak (01:15:12.322)
Yeah, well, and, you know, especially like you mentioned against this backdrop of not only an increasingly crowded market, but certain players that are able to price it, you know, below their costs. that meta display, for example, especially with that neural band and you know, that that's well below their costs. And, you know, there's a bunch of even startups who are playing a slightly different game and they're able to price it.

at or below cost. So that's the backdrop and we need to find a way to be able to still make money as a company. So we do believe in having a bit of margin that there's just, that's important to have that discipline. And at the same time, we need to have it priced competitively, dare I say even more affordably in line with what people think glasses should cost. So.

For us, it depends on which country you're buying them in. That's typically a few hundred bucks, maybe 200, 300 bucks for a nice pair of glasses. Sometimes more. Some people will pay 400 bucks for a pair of glasses, especially with a good lens in there. We always wanted to keep it around that. We felt that like 450 or 499, that was the...

That was the point of no return. We just felt that beyond that, it just feels expensive. But if you're in like the 299, 349 price point, it's a considered purchase, it's not, hopefully it's not painful. It's not perceived as a painful purchase. And so we always wanted to stay around that mark, but we're always trying to work hard to see if we can do even better. I love the...

299 price point, 249 would be even more lovely. We'd be sort of in Warby Parker territory then. So there's something nice about that.

Jason Hiner (01:17:12.834)
Yeah. Yeah.

wanna talk about the weight a little bit too because.

You said 40 grams. I heard from an eyeglass manufacturer that the targeted weight for most glasses is 35 grams. now most smart glasses, the Metare bands, the regular ones, the non-display ones, and a lot of the other competitors, they're all about 49 or 50 grams. So they're a little bit heavier than regular glasses. And then Metare band displays are like 70 grams. So they are heavier, they are noticeably heavier.

bobak (01:17:28.272)
Mm.

Jason Hiner (01:17:48.842)
number of other smart glasses right tends to be in that above 50 more closer to like 70 75 grams so so you are almost half the weight of a lot of those you are literally half about you know half the weight almost half the weight of the Meta Ray-Ban displays that also have this color display how do get there how much is that a consideration for for design as well

bobak (01:18:12.247)
It's a critical consideration for sure. mean, if we start with the use case, for us, it's about creating a vehicle that makes it possible for an AI agent to do inference and learn over your day. And so that's only possible if, number one, the glasses can be fashionable, can be beautiful enough.

to be worn all day. You should want to wear these. But then they also have to be light enough so it doesn't feel onerous on your head after a while. You're really sacrificing for this AI agent to do what it does. It should just feel like you're wearing a lot of nice pair of glasses. But the third is that the battery life should also match. If it's gonna look and weigh all day, then the battery life should also deliver all day.

Those three things are really hard to do and that's something that on the hardware side, with every product, we really challenge ourselves to try to meet that. We look at the weight and distribution of different components for us because our use case is very sparse.

know, graphical elements and it's, you know, Bluetooth only on the wireless side of things. And it's really AI inference oriented. It's not trying to show you rich graphics in front of your eyes. That means that the types of processors and sensors that we can use, and then the battery that is required to power that stuff, it's all sort of several rungs lower down on the ladder in terms of weight and power consumption and expense. So we try to take advantage of that in terms of system architecture.

which means that we can deliver something that's more cost effective, more affordable, should say, lighter, thinner, and battery lasts all day.

Jason Hiner (01:20:10.286)
Very good, all right. Final kind of technical question about Halo, which is, you mentioned that the use case, the primary feature of Halo is this using an AI agent to enhance memory, essentially.

How do you do it? You mentioned the privacy, but could you double click on the privacy factor a little bit more? How do you do that in a way that's not saving these memories in ways that others could access? You described it as their potential dystopian effects.

of that which are very, very clear, I think, to everyone. How do you accomplish that? And then what are some of the serendipitous or really delightful ways that people could experience this that you could really make?

bobak (01:21:00.708)
This is wonderful. This is a great see the light through the window. I'm to have to take a few to talk about the change. And I'm have a little bit of a time to talk about it.

Jason Hiner (01:21:05.11)
and improve life for them with that idea of having an AI agent that helps you remember things. How do you think people are gonna enjoy that? What are the aspects of that that they're gonna find really interesting that they maybe haven't considered before?

bobak (01:21:22.475)
Yeah, so, you know, maybe to start with the second part first and then I'll speak to the privacy bit, which is just foundational. know, none of it works without, you know, first principles consideration of privacy, we think. But, you know, memory, there's a lot of folks, myself included, whose memory is fallible to the degree that I forget people's names regularly.

And the more people I talk to, the more that that seems that I'm not the only one. You know, you meet someone at a meeting or a gathering and they clearly introduce themselves and you've forgotten their name five minutes later. And so, you know, whether it's small, simple things like that, whether it's the details of a conversation with a colleague and the action items of that conversation and having an agent be able to even proactively nudge you.

that, hey, you know, these were the few action items that were assigned to you, and it's now end of day, and let's make sure we have this ready for the next meeting. You know, things like this that can really help to just clear our mind of a lot of the clutter of to-dos, keep us present with people around us, and still, you know, delivering and accountable and productive in terms of the things that we need to get done. You know, those are some of the use cases, especially in the productivity realm.

But then even like go to the store and you have a conversation with your spouse or your roommate and you talk about what would be great to buy for the house and you go to the store and it's a question to your AI agent. Hey, what do I need to pick up again? That's just a memory query right there. There's all kinds of little moments throughout a day or throughout a week where

There are things said and seen that can be ingested and used later on to just be those helpful nudges that allow you to just stay present with the people you're with and it can be a little bit of a superpower for you, even those of us who forget names and struggle with that aspect.

Jason Hiner (01:23:38.446)
Could you do things like, what was the name of that Korean restaurant that I went to with my friend Jeremy last month or things like that? Interesting.

bobak (01:23:49.067)
Absolutely, absolutely. And yeah, mean, what was it? The other day I was thinking of, the name of another startup that we're thinking of partnering with. And I was blanking on the name of the startup and I was blanking on the name of the founder. And I talked to them back in August and I was searching through my Gmail and everything. And then I just remembered, this is a memory query.

Jason Hiner (01:24:18.126)
Mm.

bobak (01:24:18.666)
You know one memory query, you know, hey, what was the name of that company I was talking to back in August? They did this type of thing and the founder was, you know it was an MIT spinoff and the founder was based in this city. And you know through a vector similarity search and you know also do some queries on the graph side of things. It was able to triangulate and then surface that precise result and you know that's it so.

So it's useful in a bunch of small ways that really add up and kind of feel magical.

Jason Hiner (01:24:54.83)
Very cool. All right, lastly, I wanna give you a chance to just talk about the privacy factor, because I'm sure almost every person that I know that would consider something like this, they would have to have a really strong sense of that to feel comfortable with it.

bobak (01:24:59.841)
Yeah

bobak (01:25:09.085)
Absolutely. And, you know, I would as well. And that was actually one of the things that we considered when doing a feature like this is, is this something that we ourselves, but also our families, our children, you know, would we want them growing up in a world with technologies like this that handle privacy in this way? And, and so, you know, that really was our own intrinsic bar for doing this. The way that we handle it is when

rich media, images, audio, when they're captured by Halo or any device that we make, that is piped over Bluetooth, encrypted Bluetooth to your phone. And then there's inference done on that rich media to generate text. So the images are captioned and the audio is transcribed and the diorized. So all of it is converted into text. And then that text is embedded.

So what that means is it's then converted to strings of numbers. those numbers, sorry, just gonna clear my throat. Okay, I'll start again. So that text is then embedded and those embeddings are strings of numbers that are incomprehensible to a person accessing that stuff off of a server. But it's really designed for

the decoder of a large language model to be able to read after the fact. So if you make a query about some audio conversation you had or a visual detail of something that you saw, then that decoder could then query against the visual or the audio side embeddings to be able to then respond precisely to your question. And there's certain techniques that we apply to those embeddings to make sure that

they're secure even if there is another language model that accesses them. So whether it's leaked to a human or leaked to another language model, we really do a lot to make sure that number one, it's all encrypted on that server. And number two, there is no rich media. The original source images, the original source audio has been tossed away after inference. So none of that stuff is accessible.

bobak (01:27:30.56)
neither by us nor anyone else because it's been tossed away. And number three, the embeddings, those strings of numbers are stored in such a way that it is impossible for another human, let alone another language model to be able to make sense of them unless it's a permitted query within our platform. So there's sort of multiple steps here to make sure that nothing is leaking from your life.

you know, cause man, what a disaster that would be, right? Is if someone's able to basically view your life through your first person point of view, it's like a black mirror episode and see your images and hear your audio. And so, you know, we thought about that and we said, absolutely not. Like that is definitely not what our platform is about. And neither the intention nor the inadvertent accident that we would hope never happens. And so we've designed things so that it's just not even a possibility.

Jason Hiner (01:28:09.848)
Yeah.

Jason Hiner (01:28:28.94)
And now the app is called NOAA, N-O-A. And from my understanding of what you just said then, are you running local language model, large language models locally from the app then?

bobak (01:28:43.272)
Yes, so that is sort of been a whole endeavor of ours is getting some small models that do that inference over the rich media, the images and the audio to get those running locally on your phone. So most modern smartphones going back a couple of years, they've got an NPU and on the iOS side it's called the Apple Neural Engine, on the Android or Qualcomm side is the NPU. And so you can run a small model there.

Jason Hiner (01:28:54.03)
Yeah.

Jason Hiner (01:29:03.821)
Yeah.

Jason Hiner (01:29:08.034)
Yep.

bobak (01:29:12.836)
and that can do the inference locally without needing to go up to a server. So in a lot of our development, of course, we've been doing it up on the server, but then when we ship product out to customers, we've been holding ourselves to a standard where we want that inference happening locally on your phone so that it doesn't need to go up to the server to get processed, just for the peace of mind of keeping the rich media of your life locally about your person.

Jason Hiner (01:29:41.518)
Yeah, very good. Last question I want to have for you, Bobak, which is to unpack this a little bit. one sec, I have a meeting coming up. I just want to tell them to push it a little bit just in case we take a little bit longer on this. So yeah, give me just one moment.

bobak (01:29:53.728)
But yeah, sure, sure.

Jason Hiner (01:30:27.862)
All right, let me reframe my question here. Okay, Bobak, well, the last question that I want to explore with you is something that you and I have talked about a little bit before, and that I've been really impressed with the way you talk about the product and what you're building on social media. You don't post a lot, but...

every post that you make I've noticed is so clear, like so clearly thought through and

What I've noticed is you talk about why you're building the product that you're building, how you're building it, and what the purpose and the goals you're trying to accomplish. And that level of purposefulness has been really impressive to me as much as the products themselves of what you're building. And because the fact that there's a lot of technology that's being built today and a lot of people like you and I that are very enthusiastic about technology, but

bobak (01:31:06.303)
I if you have lots of purposes to build a new interactive accomplish and a new platform for this, you can You can apply it to the purpose and you also take advantage of the you're part of the team.

bobak (01:31:25.438)
Thank you.

Jason Hiner (01:31:30.368)
when we build technology for technology's sake or we have this sense that just by having better technology, we will automatically improve things in society, I think history has shown that that point of view can be not only...

incorrect but also at times dangerous and can have some really negative unintended consequences, collateral damage, however you want to think about it. And we've seen that, right? We've seen that with social media. Phones were built to give us these...

computers essentially in our pockets. They could do a lot of things that we used to be able to do in computers and all of a sudden they were able to do new things as well. Social media came on top of that, but the unintended consequences of what that's had on people's relationships, on the mental health of teenagers and young people and mental health of adults as well. Those things, now it's taken us a while to come to grips with and we're having to change some behaviors because of that. We have the example going back to the splitting of the

Adam, know, people did and could never have anticipated or perhaps didn't fully anticipate what would be done with that in addition to creating new forms of sustainable power, obviously using it to create weapons. That's a very extreme example, but if you think about those two examples, I think there's a large...

swath of things in between that show when we're building technology only for the sake of technology or we think that automatically it will create good outcomes for humanity, we've learned that, you know, the hard way that that's not always true. So what I've been impressed with is you've talked very clearly about why you're building and you talk about this in social media regularly, why you're building it as an open source product, you know, what the purpose is of the main features and what you're hoping will come about. And then the protections that you have in

bobak (01:32:55.582)
.

bobak (01:33:04.638)
.

the purpose of this webinar.

Jason Hiner (01:33:23.888)
thinking about what are the negative possible consequences of people using it. So before the product's even been released, you're almost open sourcing your ideas of how you've built the product and disseminating that knowledge so that the people who are interested in the product and are be part of the community with the product can have that full understanding and then in a sense kind of co-create the whole idea with you.

That's been really fascinating to me and I'd love for you to talk a little bit more about the how and why of that.

bobak (01:33:59.387)
Yeah, it's something that I care a lot about as you put it out. We all love technology. It's so interesting and especially nowadays it feels like it's moving faster and faster in its development and maturity. But

You know, I often think about technology like I do any other object in my life, which is why does it exist? You what role does it play and what purpose does it serve? And I think that we ought to be thinking about technology, whether it's a software feature or a new device in that way. And if we're considering that question, then we also ought to think about what are those unintended consequences because

we, whether consciously or not, we bring a set of values and perspectives into the development of this piece of technology. It will have an impact on the lives of the people who use it and the people who interact with the people who use it. So there's sort of these multiple degrees, these ripple effect impacts of introducing something new into the world. And on a societal level, I feel that, you know, me and...

Anyone else working on a new piece of technology has responsibility to consider that. And so you mentioned the smartphone and how inadvertently, it certainly was not a, it wasn't something that was known upfront, the impact it would have on our relationships. And especially when you pair social media and that business model together with a smartphone and just some of the addictive qualities that it could have.

That was an unintended consequence. And I think that now we have the value of hindsight there to be able to bring that learning into new devices, new experiences that we build in the future. So, you know, I try to speak openly about this because it's not only something I care about, but it's something that I hope to spark a conversation around. I just think more and more of us building in tech or thinking about tech, writing about tech, there needs to be...

bobak (01:36:09.656)
an element of our shared discourse, which is not only about what people are making as cool as that stuff is, but it perhaps more importantly needs to be the purpose and that sort of responsibility. You know, the thoughtfulness around what does, what impact does this have on the lives of the people who use it? There needs to be that layer to the discourse as well, I feel. And so I, you know, in posting about it, it's sort of my attempt to

to spark that discourse.

Jason Hiner (01:36:41.934)
So you have taken some of these ideas about why open source, and we've talked about them, why build in an open way, but also build with privacy first, build with an intention of making the audience part or the customers as well, the community part of the process. But.

Clearly, you have these talking points that you talk about regularly. I'd love for you to unpack that a little bit. What are some of the ones? Could you mention them, the things that you are regularly elevating as part of the discourse around the product and not just sort of saying, we have this cool product, nobody else is doing, ours is unique, buy ours for this. You're talking about like, our product is trying to accomplish these things. And if you align with those things,

you we'd love for you to be part of the community. So I'd love for you to mention what those things are.

bobak (01:37:44.985)
Yeah, and there's sort of a few of them. One, like you mentioned, is being open. And we think of open source as much as a principled sort of concept as it is a smart strategy, we believe. On principle, we think that we're living in a day and age, especially these technologies are becoming exponentially intelligent. We think that on principle, it's important for there to be sort of a broad-based ability for people to...

not only understand how they're designed and built, but also have the recourse to be able to modify if they so choose, the software or the hardware. So we just think that there's a justice in that. We think that that's really important in this day and age, almost like, you know, I've got young kids and so it's, you know, kids being able to gather around a bunch of Legos and be able to build something and share it with a friend, but their friend can contribute to it. They can modify it.

they can share it back again. We think that there is something really important to computing and intelligent technologies having a similar dynamic when it comes to our use at a collective level. So that's sort of on the open side of things. And that really is why, that's the big why we've embraced open source. In terms of privacy, it goes back to trust. And that's not only trust between people.

And sometimes we are a little concerned about trust worldwide in our societies and whether that's still as high as it ought to be. But it's also trust between human and machine in this age where the machine is becoming, again, exponentially intelligent. And so there are gonna be certain purveyors of that machine that

Maybe have a business model or have an approach to building it that are less than trustworthy and we think that that discernment is really really important and so we want to be people who build things that are worthy of human trust and You know open source actually also goes hand-in-hand there when you can when you can open it up It's explainable you can understand what's going on. You can modify it if you so choose We think there's an element of trust there as well that dovetails nicely

bobak (01:40:09.487)
with the privacy piece. So that's open source and privacy. But then there's maybe one more, which is human agency. And this one is sort of a little more nebulous, but no less important.

We've seen from the mobile era that whether it's the physical use of the smartphone or the use of other people in the world around you as fodder for the content engine, not valued in and of themselves, these human beings, not valued for who they are, but valued as inputs to our content and ultimately to the business model of the company that runs the content engine.

That's something that we've also tried to learn from and that's something that's about to just get put on warp speed in this new era of AI, this new era of exponential intelligence. And so we really think that it's important for human agency and nobility, for the nobility of the human being to be centered, to be elevated. That even as we're building intelligent machines, intelligent...

objects that are worn and used in different ways and that's very exciting at the same time the heart of the consideration needs to be human thriving. What does it mean for a human being to be thriving and healthy and happy and whole? You know surely there's multiple elements there. One is a feeling that they're connected with the people around them, that they're able to contribute meaningfully to their community, that they still have recourse together with their neighbors, with their friends, with their colleagues.

to advance their society together. And indeed, there's powerful research on this coming out of MIT and other places, people's ability to feel a sense of contribution and impact on the advancement of their community and their nation. So we think that that shouldn't be outsourced or stripped away from them, even as these technologies become increasingly capable. And maybe sort of another element of this, of course, that the human always feels like they are in the driver's seat.

bobak (01:42:21.977)
whether it is over your data or whether it is over the user experience and what the intelligent machine is doing with your data and how you're being networked to other people, we think it's important for the human to always feel that they're in the driver's seat of that experience. so, know, undoubtedly there's more layers to the onion that we can unpack on this, but those are a few of the principles that we think are going to be really important and that we're trying to talk

Jason Hiner (01:42:53.496)
fascinating stuff, really interested and impressed by the work that you've done and excited for you and Brilliant Labs team on your next product, Halo. So yeah, best of luck.

bobak (01:43:09.037)
Thank you so much, Jason. Great to be with you.