Bee AI: The Wearable Ambient Agent

Hey, everyone. Welcome to the Latent Space Podcast. This is Alessio, partner and CTO at Decibel Partners, and I'm joined by my co-host, Swyx, founder of Smol AI.

Hey, and today we are very honored to have in the studio Maria and Ethan from Bee. Welcome.

Hi. Hi.

Well, uh-

Thank you for having us.

Yeah. And, uh, you are, I think, the first hardware founders we've had on the podcast. I've been looking to have, uh, had a hardware founder... Like, a wearable hardware. Like, yeah, a wearable hardware founder for, for a while.

I think we're gonna have, uh, two or three of them this year. Um, and you're the ones that I wear every day, so thank you for making Bee.

Thank you.

Oh, thank you for, for, for all the, all the feedback and the usage.

Yeah, you know, I've been, I've been a big fan. You were the speaker gift for AI Engineering World's Fair. And, uh, let's start from the beginning. What is Bee Computer?

Bee Computer is a personal AI system, so you can think of it as AI living alongside you in first person, so it can kind of capture your in real life context. So with that understanding, can help you in significant ways.

You know, the obvious one is memory, but that's, that's really just the base kinda use case. So recalling and, and reflective. I, I know, Swyx, that you, you like the idea of journaling, but you don't, but still have some, some kind of reflective summary of what you experienced in real life.

But it's also about just having, like, the whole context of a, a human being and understanding. You know, giving the machine the ability to understand, like, what's going on in your life, your attitudes, your desires, specifics about your preferences, so that not only can it help you with recall, but then anything that you need it to do, it already knows.

Like, if you think about, like, somebody who you've worked with or lived with for a long time, they just know kinda without having to ask you what you would want. It's clear that, like, that is the future, that personal AI, like, is just gonna be very...

You know, that AI is just so much more valuable with personal context.

I would say that one of the things that we are really passionate is really understanding this personal context because we'll make the AI more useful. Think about, like, a best friend that know you so well. That's one of the things that we are seeing from the user.

They're using from a companion standpoint or professional use cases. There are many ways to use Bee, but companionship and professional are the ones that we are seeing now more.

Yeah. It feels so dry to talk about use cases.

Yeah. Well-

Yeah, it's, like, really, like, investor question, like what are your use cases?

Yeah, we're just like-

So we are really here trained.

We've been so broken and trained

To say that.

But, I mean, like, on, on, on the base case, it's just like, don't you want your AI to know everything you've said-

Mm-hmm

... and, like, everywhere you've been? Like, wouldn't you want that? Because-

Yeah, and don't say there and repeat every time. Like, "Oh, this is what I like." You already know that, um, and you do things for me based on that.

Yeah.

That I think is really cool.

Great. Um, do you wanna jump into a demo? Do you have any other final questions?

I, I wanna maybe just cover the origin story.

Yeah.

Okay.

Just how did you two meet? What was the-- Was this the first idea you started working on? Was there something else before?

I can start. So Ethan and I, we know each other from six years now. He had a company called Squad, and before that was called Olabot, and was a personal AI. Uh-

Yeah, I, I should-- Yeah.

So maybe you should start this one. But yeah, that's how I know Ethan. Like, he was pivoting from personal AI to Squad, and that was a co-watching with friends product. I had experience working with TikTok and video content, so I helped the pivoting, and we launched Squad, and was really successful.

And at the end, uh, the founders decided to sell that to Twitter, now X. So both of us, we joined X. We launched Twitter Spaces. We launched many other products. And yeah, till then, we basically continued to work together to the start of Bee.

The interesting thing is, like, this isn't the first attempt at personal AI. In 2016, when I started my first company, it started out as a personal AI company. This is before transformers, no BERT even, like, just RNNs. You couldn't really do any-

Mm-hmm

... convincing dialogue at all. I met Esther, who was my previous co-founder, and we were both really interested in the idea of, like, having a, a machine kind of model or understand a dynamic human. We wanted to make personal AI.

This was, like, more geared towards-- 'cause we had obviously much limited tools, more geared towards, like, younger people. So I don't know if you remember in 2016, there was, like, a brief chatbot boom. It was way premature, but it was-

D?

... when Zuckerberg went up on F8 and yeah, M, and, like, the Messenger platform, people were like, "Oh, bots are gonna replace apps." It was, like, for about six months, and then everybody realized, man, these things are, are terrible and, like, they're not replacing apps.

But it was at that time that we got excited, and we're like, we tried to make this like, oh, teach the AI about you. So it was just an app that you kinda chatted with, and it would ask you questions and then, like, give you some feedback.

But-

Well, Hugging Face first version was launched at the same time.

Yeah, we started it-- We, we started out of the same office as Hugging Face 'cause Betaworks was our investor, so they had a thing called Bot Camp. Betaworks is, like, a really cool VC because they invest in out-there things.

They're, like, way ahead of everybody else. And, like, back then it was-- They had something called Bot Camp. They took six companies, and it was us and Hugging Face, and then I think the other four, I'm pretty sure, are dead.

But, um- And Hugging Face was the one that really got, you know, got the return.

Yeah, that was the one.

I mean, thirty percent success rate is pretty good.

Yeah.

Yeah. Um, but yeah, when we-- It was, it was like, it was just, uh, the two founders, Clement and Julian.

Yeah, we were doing, like, an AI companion at the beginning.

Yeah, it was a, it was a chat app for teenagers.

Yeah.

A lot of people don't know that Hugging Face was like, "Hey, friend, how was school? Let's trade selfies." But then, um, you know, they built the Transformers library, I believe, to help them make their chat app better.

Yeah.

And then they open sourced and it was like it blew up and, like, they're like, "Oh, maybe this is the opportunity," and now they're Hugging Face. But anyway, like, we were obsessed with it at that time, but then it was clear- That there's some people who really love chatting and, and like answering questions, but it's like a lot of work, like, just to kind of manually-

Yeah

... teach, like, all these things about you to an AI.

Yeah, there were some people that were super passionate. Uh, for example, teenagers, they really like, for example, to speak about themself a lot, so they will reply to a lot of questions and speak about them. But most of the people, they don't really want to spend time speaking.

And, you know, it was hard to like really bring the value with it. We had like sentence similarity and stuff and could try and do-- But it was like it was a little-- It was premature with the, the technology at the time.

And, uh, so we, we pivoted, we went through YC, and long story, but like we pivoted to consumer video and, and that kind of went really viral and got a lot of usage quickly. And then we ended up selling it to Twitter, worked there and left before Elon.

Not related to Elon, but left Twitter. And then-

I, I should mention, this is the famous, uh, time when... Well, when, when Elon was-- just came in. This was like Esther was the famous-

Yes. Yes

... product manager who was left in the-

My co-found-

Yeah

... my former co-founder. She, um-

Sleeping bag

... she was the, the sleep warrior.

She's one of our investors as well.

Yeah, yeah, she invest. She stayed. We had left by that point.

She very stayed.

She very stayed.

She's famous for staying.

Yeah. But later, later left or got, I think, um-

Laid off

... laid off.

Yeah.

I think the whole product team got laid off. She was a product manager, director. But yeah, like we left before that and, um, then we're like, "Oh my God, things are different now."

Yeah.

You know, I think this is-- We, we really started working on it again right before ChatGPT came out. But we had an app version, and we kind of were trying different things around it. And then, you know, ultimately, it was clear that, like there were some limitations.

We can go on, like the-- A good question to ask any wearable company is like, "Why isn't this an app?"

Yes.

Yeah.

Um, because like-

Because we tried the app at the beginning.

Yeah. Like the idea that it could be more of a... And Bee comes from ambient. So like-

Mm-hmm

... if it was more kind of just around you all the time and less about you having to go open the app and do the effort to like enter in data, that led us down the path of, of hardware because the sensors on this are, are microphones, so it's capturing and understanding audio.

We started actually our first hardware with, with a vision component too, and we can talk about why-

That's very cool

... why we're not doing that right now. But if you wanted to like have like continuous understanding audio with your phone, it would monopolize your microphone.

Mm-hmm. Yep.

It would get interrupted by calls, and you'd have to remember to turn it on. And like that little bit of friction is actually like a substantial barrier to like the experience of it just being with you all the time and like living alongside you.

Mm-hmm.

And so I think that that's like the key reason it's not an app. And in fact, we do have Apple Watch support. So anybody who has a Watch, Apple Watch, can use it right away without buying any hardware-

Hmm

... 'cause we worked really hard to make a version for the watch that can run in the background, not super drain your battery. But even with the watch, there's still friction because you have to remember to turn it on, and it still gets interrupted if somebody calls you, and you have to remember to-- We'd send a notification, but you still have to go back and turn it on because of just the way watchOS works.

One of the things that we are seeing from our Apple Watch users, like I love the Apple Watch integration. One of the things that we are seeing is that people, they start using it from Apple Watch, and after a couple of days, they buy the Bee because they just like to wear it.

Yeah, we're seeing, yeah, we're seeing-

Um, like that's something that like we're learning-

Oh

... and it's really cool.

Yeah, I mean, we-

Yeah

... we, we-- I think like fundamentally, we, we like to think that like a personal AI is like the mission, and, and it's more about like the understanding, connecting the dots, making use of the data to provide some value.

And the hardware is like the ears of the AI. It's not like integrating like the incoming sensor data and that, that's really what we, we focus on. And like the hardware is, you know, if we can do it well and have a great experience on the Apple Watch, like that, that's just great.

I mean, but there's just some platform restrictions that like existing hardware makes it hard to provide that experience.

Yeah. What do people do in like two, three days that then convinces them to buy the product? This feels like a product where like after you use it for a while, you have enough data to start to get a lot of insights.

Yeah, which is-

But sounds like maybe there's also like a short term.

From the Apple Watch users, I believe that because every time that you receive a call, after they need to go back to Bee and open it again, or for example, every day they need to charge Apple Watch and reminds them to open the app every day.

Oh, yeah.

They feel like, "Okay, maybe this is too much work. I just want to wear the Bee and just keep it open, and that's it, and I don't need to think about it."

I think they see the kind of potential of it just from the watch, 'cause even if you wear it a day, like we send a, a summary notification at the end of the day about like just key things that happened to you in your day.

And like, I didn't even think like-- I'm not like a journaling type person or like, 'cause like, "Oh, I just live the day. Why do I need to like think about it?" But like, it's actually pretty-- Sometimes I'm surprised how interesting it is to me just to kind of be like, "Oh yeah, that..."

And how it kind of fits together, and I think that's like just something people get immediately with the watch, but they're like, "Oh, I'd like an easier way to do this."

It's, uh, surprise- surprising 'cause I only, I only know about the hardware.

Yeah.

Mm-hmm.

Um, but I use the watch as like a backup for when I-

Yeah

... don't have the hardware. I feel like the-- Because now you're beamforming and all that-

Yeah

... this is significantly better.

Yeah, that's the other thing.

Oh, thanks.

We have way, way more control over... Like the, the Apple Watch, you're limited in like you don't-- you can't set the gain. You can't change the sample rate. You c- There's just a very limited framework support for doing anything with audio, whereas if you control it, then you can kind of optimize it for your use case.

The Apple Watch isn't meant to be kind of recording this. And we can talk when we get to the part about audio, why it's so hard. This is like audio on the hardest level because you don't know. It has to work in all environments, or you try and make it work as best as it can.

Like this environment's very great. We're in a studio. But, you know, afterwards, at dinner, in a restaurant, it's a totally different audio environment.

Yeah.

And there's a lot of challenges with that. And having really good Source audio helps, but then there's a lot with the machine learning that still is, you know, has to be done to try and account. 'Cause, like, you can tune something for one environment or another, but it'll make one good and one bad-

Mm-hmm. Yeah

... and, like, making something that's flexible enough is, um, really challenging.

Do we wanna do a demo just to set the stage-

Yeah

... and then we can kind of talk about-

Yeah, I think we can do, like, a walkthrough-

... how it comes together

... the, the prod.

Yeah, sure. So, um, I think we said I should, uh-

So for listeners, uh, we'll be switching to video. It w- that will superimpose on, on to this video. If you wanna see, go, go to our YouTube, like and subscribe as always.

Yeah. And buy the Bee.

Yes, and buy the Bee.

While you wait- ... for the video to come up-

While you wait, exactly

... buy the Bee.

'Cause it's gonna take a while.

Hopefully it doesn't take long. Uh-

Maybe you should have a discount code just for-

Yeah. No, we should

... the listeners.

Sure. If you want to offer it, I'll take it.

Yeah. Uh-

We can put it on there

... yeah, well, discount code SWIX. Uh

Oh, shit. Okay. Yeah, there you go.

An important thing to mention also is that the hardware is meant to work with the phone and, like, I think, you know, if you, if you look at Rabbit or, or Humane, they're trying to create, like, a new hardware platform.

Mm.

We think that the phone's just so dominant, and it will be until we have the next generation, which is not gonna be for five, you know, maybe some Orion-type glasses that are cheap enough and, like, light enough. Like, that's gonna take a long time before with the phone, rather than trying to just, like, replace it.

So in the app, we have a summary of your days, but at the top it's kind of what's going on now, and that's updating continuously. So right now it's saying, I'm discussing, you know, the development of, you know, personal AI, and that's just kind of the ongoing conversation.

And then we give you a readable form that's like little kind of segments of what's the important parts of the conversations. We do speaker identification, which is really important because you don't want your personal AI thinking-

Right

... you said something and attributing it to you when it was just somebody else in the conversation. So you can also teach it other people's voices, so, like, if some, you know, somebody close to you, so it can start to understand your relationships-

Mm-hmm

... a little better. And then we do conversation end pointing, which is kind of like a, a, a task that didn't even exist before-

Mm-hmm. Right

... like, 'cause nobody needed to do this. But, like, if you had somebody's whole day, how do you, like, break it into logical pieces? And so we use, like, not just voice activity, but other signals to try and split it up because conversations are a little fuzzy.

Mm-hmm.

They can, like, lead into one, can start to the next, so also, like, the semantic content of it. When a conversation ends, we, we, we run it through larger models to try and get a, a better, you know, sense of the actual, what was said and then summarize it, provide, um, key points, what was the general atmosphere and tone of the, the conversation and potential action items that might have come of that.

But then at the end of the day, we give you, like, a summary of all your day and, and where you were and just kind of like a step-by-step walkthrough of, of what happened and what were the key points.

That's kind of just like the base capture layer. So, like, if you just wanna get a kind of glimpse or, or recall or reflect, that's there. But really the key is, like, all of this is now, like, being inferenced on to generate personal context about you.

So we generate key items known to be true about you and that you can be-- You know, there's a human in the loop aspect is like you can-

Mm-hmm

... you have visibility-

Right, right

... into that, and you can... You know, I have a lot of facts about technology 'cause that's basically what I talk about all the time.

Right.

But I, I, I do have some hobbies- ... that show up. And then, like, how do you put use to this context? So I, I kind of like measure my day now and just like what is my token output of the day, you know?

Like, like as a human, how much information do I produce? And it's kind of measured in tokens, and it turns out it's like around two hundred thousand or so a day. But so in the recall case, we have, um, a chat interface, but the key here is on the recall of it.

Like, you know, how do you... You know, I probably have fifty million tokens of personal context and, like, how to make sense of that, make it useful. So I can ask simple, like, uh, recall questions, like details about the trip I was on to Taiwan, where I recently were with our manufacturer.

And, um, in real time, like, it will-- You know, it has various capabilities such as searching through your, your memories, but then also being able to search the web or look at my calendar. We have integrations with Gmail and Calendar, so, like, connecting the dots between the in real life and the digital life.

And, you know, I just asked it about my Taiwan trip, and it k- kind of gives me the, the breakdown of the details, what happened, the issues we had around, you know, certain manufacturing problems, and it, and it goes back and references the conversation.

So I can, I can go back to the source.

Yeah, not just the conversation, as well, the integrations.

Yeah.

So we have as well Gmail and Google Calendar, so if there is something there-

Yeah

... that was useful to have more context, we, we can see that.

So, like, and it can-- I never use the word agentic 'cause it's, it's cringe. But, like, it can-

Like an agency.

Yeah.

Yeah.

It, it can search through, you know, if I, if I'm brainstorming about something that spans across, like search through my conversation, search the email, look at the calendar, and then depending on what's needed, then synthesize, you know, something with all that context.

I love that you did the Spotify Wrapped. That was pretty cool.

Yeah, like-

Yeah

... one, one thing I did was just, like, make a Spotify Wrapped for my 2024, like, of my life.

You can do that?

Yeah.

Yeah.

You can.

Wait.

Yeah. I thought it was crazy.

Let's hear a Spotify Wrapped.

Make a Spotify Wrapped-

For my life in 2024

... of my 2024. Yeah.

So it, it's, like, surprisingly good. Um, it, like, kind of, uh, like, gave metrics. Oh, it's like you visited three countries. You shipped-

Mm

... you know, X many beta devices, and that's kind of more reflective.

Yeah, it gives a lot of personal insights and reflection points.

Yeah. That's-

Um

... that's, that's fascinating. So that's the demo. Um-

Well, we have-- We can show something that's in beta. I don't know if we wanna sh- do it. I don't-

Do we wanna show something that's beta?

Well-

Do it, and then we can cut if it-

Yeah.

Yeah. So-

Good luck.

So like the, the, the, the vision is also like not just about like AI being with you and like just passively understanding you through living your experience, but also then like it proactively suggesting things to you, like at the appropriate time.

So like not just pull, but, but kind of it can step in and, and suggest things to you. So you know, one integration we have that, uh, is in beta is with WhatsApp. Maria's asking for a recommendation for an Italian restaurant.

Yeah. Would you like me to look up some highly rated Italian restaurants nearby and send her a suggestion?

So what I did, I just sent to Ethan a message through WhatsApp in his own personal phone.

Yeah. So, so basically Bee is like watching all my incoming notifications, and if it meets two criteria, like is it important enough for me to raise a, a suggestion to the user, and then is there something I could potentially help with?

So this is where the actions come into place.

Mm-hmm.

So because Maria's my co-founder, and because it was like a restaurant recommendation, something that it could probably help with-

Mm-hmm

... it proposed that to me. And then I can, through either the chat, and we have another kind of push-to-talk walkie-talkie style button.

Hmm.

It's actually a multipurpose button to like toggle it on or off, but also if you push to hold, you can talk. So I can say, "Yes, uh, find one and send it to her on WhatsApp." Is, uh, an Android cloud phone, so it's, uh, going to be able to...

you know, that has access to all my accounts. So we're gonna abstract this away, and the execution environment's not really important.

Yeah.

But like we can go into technically why Android is actually a pretty good one right now. But you know, it's searching for Italian restaurants. You know, and we don't have to watch this. I could be, you know, have my ear- AirPods in and in my pocket.

You know, it's gonna go to WhatsApp, gonna find Maria's thread, send her the response, and then, and then let us know. Oh my God.

Well, what's again, it's a Je- I mean, an Italian restaurant.

Yeah. What did it choose? What did it choose? It's easy to say-

Italian's are like a real Italian is hard to please.

It's easy-- Exactly. It's easy to say- I doubt it's... I don't know what-

But-

For, for the record, since we have Italians, uh, best Italian restaurant in SF.

Oh my God. I still don't have one.

Name it.

What?

No. Um, I don't...

Successfully found and shared-

Let's see, let's see what the AI says

... with Maria via WhatsApp.

Bottega.

Bottega. Have you been to Bottega?

Have you been to Bottega? How is it?

Four stars-

It's fine

... distance.

I've been to one called like Norcina, I think.

Hmm.

It was good.

Bottega's on Valencia Street. It's fine.

Yeah.

The pizza is not good.

It's not good?

Some of the pastas are good, but-

You know that the people say that. I'm sorry to interrupt you, Jacob.

Sorry.

But there is like this Delfina that here everybody's like-

Uh-huh

... "Oh, Pizzeria Delfina is amazing," and like-

Overrated

... this is not... I don't know.

It's great. It's great.

To North- North Beach Cafe.

No.

That, that place you took us with Michele last time.

Vega.

Oh.

The guy at Vega, Giuseppe, he's Italian.

Which one is it?

It's in Bernal Heights.

Ugh.

He's nice.

No, I don't know that one.

He's a nice-

What's the name of the place?

Vega.

Vega.

Vega?

Okay, cool, cool. We got the name.

Vega, but it's not Vega. Is it Italian? What does it mean in Italian? Vega niente.

It doesn't mean to me.

The last thing I'll mention is also the-- you have wake word w- with detection.

Yeah.

So the, the phone can be off, and you can, you can just like say, "Hey," something. Mine is Alfred.

Yeah, yeah.

So I say, "Hey, Alfred," and then it just does the thing.

Mm-hmm.

Yeah.

Just being able to have wake words, like enables-

Mm-hmm

... some form of sort of voice agent, agent features that even ChatGPT can never have because they don't have the recording layer.

Yeah, I think we have some other ideas like even beyond wake words, but I think that it's interesting to see how, how people use the voice side of it and, um, I think that, you know, we're gonna see a lot of innovation around hardware and stuff, but I think the, the real core is like being able to do something useful with the personal context 'cause it does, like...

The thing that like you, you always had the ability to capture everything, right? Like you've-- we've always had recorders, camcorders-

Mm-hmm

... body cameras, stuff like that. But like the-- what's different now is like we can actually make sense and find the important parts in all of that context.

Yeah, so and then one last thing, I'm, I'm just doing this, this for you, uh, is you also have an API, which, uh, I think I'm the first developer against this.

Yeah. Yeah.

'Cause I had to build my own episode, like-

We need to, we need to hire a, a developer advocate, um.

Or, or just, yeah, hire AI engineers.

Yeah.

The point is that you should be able to program your own assistant.

Mm-hmm.

And I tried Omi, the former Friends, the knockoff Friend, and then, you know, real Friend doesn't have an API, and then Limitless also doesn't have an API. So like I think it's very important to own your data, to be able to reprocess your audio maybe, uh, although like by default you do not store audio.

Yeah.

And then also just to like, yeah, to, to, to do any corrections. Like there, there's no way that my needs y- can be fully matched by you.

Yeah.

So I think the API's very important.

Yeah. And I mean, as we're-- I've always been a consumer of APIs, you know, in all my products. And so, um-

We are API enjoyers in this house.

Yeah. Yeah. I am. It's very frustrating, like when you have to go like build a scraper or, you know, like, um, but yeah, it's, it's for sure.

Yeah. So like this, this, th- th- this is all combination of like you have my location, my calendar, my invi... Like it's, it's, it really is-

Yeah

... for me the, the sort of personal, yeah.

And is the API just to write into it or to like have it take action on external systems?

Yeah, we're expanding it. It's right now read-only. In the future, very soon, when the actions are more generally available, it'll be fully supported in the API.

Mm-hmm. Nice. I'll buy one after the episode. The API thing to me is the most interesting.

Hey.

Um-

We do have like real-time APIs, so like you can even connect a socket and like connect it to whatever you want it to take actions with.

Yeah. It's too, too smart for me.

Yeah. I'm, uh-- Yeah, I think like when I look at these apps, and I'm... I mean, there's so many of these products-

Yeah

... we launch, it's like, it's great that I can go on this app and do things, but like most of my work and like personal life is like managed somewhere else.

Yeah.

Mm-hmm. Mm-hmm.

So being able to like plug into it-

Integrate that

... it's nice.

Yeah.

I have a bunch of more maybe human questions-

Sure

... that I think maybe people might have. One, is it good to have instant replay for any argument that you have?

Yeah.

I can imagine arguing with my wife about something and, you know, there's these commercials now where, uh, it's basically like two people arguing, and they're like, they can throw a flag like in football and have an instant replay-

Okay.

Mm-hmm

... of the conversation.

Yeah.

I feel like this is similar, where it's almost like-

People cannot really argue anymore-

Mm-hmm

... or like lie to each other- ... because in a world in which everybody adopts this. I don't know if you thought about it, and also like how the lies-- You know, all of us tell lies, right? How do you distinguish between when I'm s- There's gonna be sometimes things that contradict each other, because I might say something publicly, and I might think something really that I tell someone else.

How do you handle that when you think about building a product like this?

I would say that I like the fact that Bee is an objective point of view.

Mm-hmm.

So I don't care too much about the lies, but I care more about the fact that can help me to understand what happened and the emotions in a really objective way, like really like critical and objective way. And if you think about humans, they have so many emotions, and sometimes something that happened to me, like, I don't know, I will feel like really upset about it-

Mm-hmm

... or really angry or really emotional, but the AI doesn't have those emotions. It can read the conversation, understand what happened, and be objective. And I think that level of support is the one that I really like more instead of like, "Oh, did this guy tell, told me a lie?"

Mm-hmm.

I feel like that's not exactly like what I find curious for me-

Yeah

... the in terms of opportunity.

Is the Bee going to interject in real time? Say I'm arguing with somebody, the Bee's like, "Hey, look, no, you're wrong."

Well-

"That, that person actually said..."

The proactivity is something we're very interested. Maybe not for like specifically-

Right. Yeah

... for like selling arguments, but more for like-- And I think that a lot of the challenge here is, um, you know, you need really good reasoning to-

Mm-hmm

... kind of pull that off because you don't want it just constantly interjecting 'cause that would be super annoying.

Mm-hmm.

And you don't want it to miss things that it should be interjecting. So like it'd be kind of a hard task even for a human to be like, just come in at the right times when it's appropriate. Like-

Yeah.

It would take the-- And, you know, with the personal context, it's gonna be a lot better 'cause like if somebody knows about you, but even still, it requires really good reasoning to like not be too much or too little and just right.

And the second part about, uh, well, like some things, you know, you say something to somebody else, but after I change my mind, I send something. Like it's every time I have like different type of, uh, conversation and data about me.

I think that's something that I found really fascinating. One of the things that we are learning is that indeed humans, they evolve over time.

Mm-hmm.

So for us, one of the challenges is actually understand like is this a real fact?

Right.

And so far, what we do is we give, you know, to the u- we have the human in the loop that can say like, "Yes, this is true. This is not," or they can add it their own fact.

For sure in the future we want to have all of that automatized inside of the product.

But I mean, I think your, your question kinda hits on, and I know that we'll, we'll talk about privacy, but also just like if you have some, some memory and you wanna confirm it with somebody else, that's, that's one thing.

But it's for sure going to be true that in the future, like not even that far into the future, that it's just going to be kinda normalized, and we're kind of in a transitional period now. And I think it's like one of the key things that is for us to, to kinda navigate that and make sure we're like thinking of all the consequences and, and how to, you know, make the right choices in, in the way that, that everything's designed.

And so like it's, it's more beneficial than it could be harmful. But it's just too valuable for your AI to understand you. And so if it's like Meta Ray-Bans or the Google As-Astra, um, I think it's just people are gonna be more used to it, so people's behaviors and expectations will change.

Whether that's like, you know, something that is going to happen now or in five years, it's probably in that range. And so, like I think we kind of adapt to new technologies all the time. Like when the Ring cameras came out, that was kinda quite controversial.

It's like, but now it's kinda-- people just understand that a lot of people have cameras on their doors and so I think that-

Yeah, we're in a transitional period for sure.

I will press on the privacy thing because that is the number one thing that everyone talks about. Obviously, I think in Silicon Valley, people are a little bit more tech forward, experimental, whatever, but you want to go mainstream.

You want to sell to consumers, and we have to worry about this stuff. Baseline question, the hardest version of this is law. There are one-party consent states where this is perfectly legal, then there are two-party consent states where they're not.

Um, what have you come around to this on?

Yeah, so the EU is, is totally different, uh, regulatory environment. But in the US, it's basically on a state-by-state level. Like in Nevada, it's single party. In California, it's two party. But it's kind of untested. You know, it's different laws, whether it's a phone call, whether it's in person.

In a, in a state like California, like anytime you're in public, there's no consent comes into play 'cause the expectation of privacy is that you're in public. But we process the audio and nothing is persisted, and then it's summarized with the speaker identification focusing on the user.

Now, it's kind of untested on a legal, and, and I'm not a lawyer, but does that constitute the same as like a, a recording? So, you know, it's, it's kind of a, a gray area and untested in, in law right now.

I think that the bigger question is, you know, because like if you had your Ray-Ban on and we're recording, then you have a video of something that happened, and that's different than kind of having like an AI give you a summary that's focused on you that's not really capturing anybody's voice.

You know, I think the bigger question is regardless of the legal status, like what is the, the ethical kind of situation with that? Because even in Nevada that we're, we're-- or many other US states where you can record everything and you don't have to have consent, is it still like the right thing to do?

The way we think about it is, is that, you know, we take a lot of precautions to kind of not capture personal information of people around, um, both through the speaker identification, through, through the pipeline, and then the prompts and, and the way we store the information to be kind of really focused on the user.

Now, I-- we know that's not gonna like satisfy a lot of people, but I think if you do try it and wear it, it's very hard for me to see anything, like if somebody was wearing a Bee around me, that I would ever object to it.

Captured about me as like a third party to it. And like I said, like, we're in this transitional period where the expectation will just be more normalized that it's, it's like an AI. It's not capturing, you know, a full audio recording of what you said, and it's, it's-- everything is fully geared towards helping the person kinda understand their, their state and providing valuable information to them, not about, like, logging details about people they encounter.

You know, I've had the same question also with the Zoom meeting transcribers-

Yeah

... thing. I think there's kinda like the personal impact that there's a Fireflies AI recorder.

Yeah.

I just know that it's being recorded. It's not like I, I don't know if I'm gonna say anything different, but, like, intrinsically you kinda feel because it's not pervasive, and I'm curious, especially, like, in your investor meetings, do people feel differently?

Like, have you had people ask you to, like, turn it off, like, in a business meeting to not record? I'm curious if you've run into any of, of these behaviors.

You know what's funny? On my end, I wear it all the time. I take my coffee blue bottle with it, or I work with it, like, obviously I'm working it, so I wear it all the time. And so far, I don't think anybody asked me to turn it off.

I'm not sure if because they were really friendly with me that they know that I'm working on it, but nobody really cared.

It's 'cause you live in SF.

Actually, I've, I've been in Italy as well.

Uh-huh.

And Italy is super privacy concerned. Like, Europe is super privacy concerned. And again, they're nothing. Like, it's-- I don't know.

Yeah.

Uh, that for me was interesting.

I think, yeah, nobody's ever asked me to turn it off, even after giving them full demos and, and disclosing. I think that some people ha- have said, "Well, my," you know, "in a, in a personal relationship, my partner initially was, like, kinda uncomfortable about it."

We heard that from a few users, and that was, like, more in just, like, a personal relationship, um, situation. And, um, the other big one is people are like, "I do like it, but I cannot wear this at work-

Mm-hmm.

Oh, yeah.

Yes. Yeah

Yeah

... 'cause, like, I think I will get in trouble based on policies" or, like, you know, if you're, you're wearing it inside a, a research lab or something-

Yeah

... where you're working on things that are kinda sensitive that, like-- They're like, you know-- So we're adding certain features like, like geofencing, just, like, at this location, it's just never active. And even, like, concept fencing, so you can be like, if these topics come up- ...

then, like, n- don't-

Yeah, yeah

... no capturing of that.

Uh, w-- I mean, I've often off- often actually explained to it the other way, where maybe you only want it at work.

Yeah.

So you never take it from work, and it's, it's just a work device, just like your Zoom meeting recorder is your work device.

Yeah, professionals have been a big early adopter segment. You say in San Francisco, but, like, we have out there a daily shipment of, like, over 100. If you go look at the addresses and like-

Yeah, who's buying these things?

They're, they're, they're in, like, Texas, I think is our biggest state, and Florida. Like, just the biggest states. Like, a lot of professionals, like, who talk for-- You know, and we didn't go out to build it for that use case.

But, like, I think there is a lot of demand for, like, kinda white-collar people who talk for a living, and I think we're just starting to talk with them. I think they just wanna be able to, like, improve their performance around, you know, like understand what they were doing-

Yeah

... and, and improve, so.

How do you think about Gong.io? Some of these, like, for example, sales training thing where, like, you put on a sales call, and then it coaches you through-- They're more verticalized versus having kinda like a more horizontal platform.

Yeah. I, I, I am not super f-familiar with the space 'cause like I said, we weren't, like-

Right. Yeah, yeah. It's, um-

... it was kind of a surprise to us, but I think that, that those are interesting. I've seen there's, there's a, there's a bunch of them now, right? 'Cause, like, it, it kinda makes sense. I'm terrible at sales, so, like, I could probably use one.

But, like, it's not my job fundamentally, but- But, um, yeah, I think maybe it's, you know, in, in a little f- We heard also, like, people with restaurants, like, if they're able to understand, like, if they're, they're doing well, um-

Yeah, but in general, I think a lot of people, they like to have the double check of, "Did I do this well?"

Mm-hmm.

Or, "Can you suggest me how I can do better?" We had a user that was saying to us that used for interviews, job interviews. So use Bee, and after ask Bee, "Oh, actually, how do you think my interview went?

What I should do better?" And I like that. I'm like, oh, that's actually like a personal coach in a way.

Yeah. But I guess the question is, like, do you wanna build all of those use cases, or do you see Bee as more like a platform where somebody's gonna build, like, you know, the sales coach that connects to Bee so that you're kind of the data feed into it?

I don't think it's just like a data feed, more like a understanding kind of engine, and, like, definitely in the future, having third parties through the API and building out for all the different use cases is something that we, we want to do.

But the, like, initial case we're trying to do is, like, build that layer for all that to work. And, um, you know, we're not trying to build all those verticals 'cause no startup could do that well. But I think that it's really been quite fascinating to see, like, you know-- And I-- I've done consumer for a long time.

Consumer is very hard to predict, like, what's gonna be, like, the thing that's the killer feature. And so, I mean, we really believe that it's the future, but we don't know, like, what exactly, like, process it will take to really gain mass adoption.

The, the killer consumer feature is whatever Nikita Bier does.

Yeah. Yeah.

Social app for teens.

Yeah, well, I like Nikita, but, um, you know, he's, he's good at building bootstrap companies and getting them very viral and-

And then selling them, and then they shut down. Okay, so, uh, you just came back from CES?

Yeah. Crazy.

Yeah, tell us.

Uh, it was my first time in Vegas and first time CES. Both of them were overwhelming.

First of all, did you, did you feel like you had to do it because you're in consumer hardware?

Uh, then we decided to be there and to have a lot of partners and media meetings, but we didn't have our own booth. So we decided to, to skip that. But we decided to be there and have a presence there, even just us, and speak with people-

It's very hard to stand out

... that was needed. Yeah, I think, you know, it depends what type of booth you have. I think if you can prepare like a really cool booth-

Have you been to CES?

I think it can be pretty cool.

It's, it's massive.

It's huge. Yeah.

It's like eighty, ninety thousand people across The Venetian and the convention center, and it's- To me, I always wanted to go just, like, uh, even-

Yeah, you were the one that was like-

A lot-

Oh. I thought it was your, I thought it was your idea.

I, I, I always wanted to go-

No, he was the-

... just as a, like, just as a fan of-

I think... Yeah, you wanted to go anyways

... 'cause like, I w- growing up, I think CES, like, kind of peaked for a while, and it was like, "Oh, I wanna go there. That's where all the cool, like, gadgets, everything is, is launched."

Yeah, now it's, like, smart fish and like-

Yeah

... you know, vac- vacuum that picks up socks.

There are... Exactly, there are a lot of cool vacuums.

Oh, they love-

Yeah

... they love the Roombas that pick up socks.

And, and pet tech.

Yeah, yeah.

For pet and dog stuff. Um-

Yeah, there's a lot of, like, robot stuff that-

New TVs, new cars they never ship.

Yeah.

Yeah. I'm thinking, like, last year, this time last year was when Rabbit-

Mm-hmm

... and Humane-

Yes

... launched at CES.

Yes.

And Rabbit kind of won CES.

Yeah.

And now this year, no wearables except for you guys.

The, it's funny 'cause, like, it's, it's obviously, it's AI everything.

Yeah.

Like, every single pro-

Yeah

... like-

With brush, with AI.

Yeah.

Vacuums with AI. Everything with AI.

Yeah, yeah, we, like, hair blow- l- literally a hair dryer with AI, um, we saw. Um-

Yeah, that was cool

... but I think that, like, yeah, we didn't... Another kind of difference, like, uh, around our... Like, we didn't wanna do, like, a big over-hypey promised kind of Rabbit launch-

Mm

... 'cause I mean, they did it. I, hats off to them, like, on the presentation and, and everything, obviously. But like, you know, we, we wanna let the product kind of speak for itself, and, like, get it out there.

And I think we're really happy. We, we got some very good interest from, from media and, and some of-

Yeah

... the partners there. So, like, it was... I think it was definitely worth going. I would say, like, if you're in hardware, it's just kind of how you make use of it. Like, I think to do it like a big Rabbit style or to have a, a huge show on there, like, you need to plan that six months in advance, and, uh, it's very expensive.

But, like, if you, you know, go there, there's, everybody's there. All the media's there. There's a lot of, um, some pre-show events that it's just great to talk to people in the industry. Also-

Yeah, I met lots of them

... all the manufacturer suppliers are there, so we learned about some really cool stuff that we might... Like, we met with, uh, somebody. They have, like, thermal energy capture, and it's like, "Oh, could you maybe not need to charge it?"

'Cause they have, like, a thermal that can capture your body heat. And, um-

What?

Yeah, they're here. They're actually here in, uh, in Palo Alto. They have, like, a f, like, a, a, a Fitbit thing that you don't have to charge-

Like, on paper-

... 'cause it works off your body heat

... does the power you can get from that... What's the power draw for this thing and-

It's more than you could get from the body heat-

Yeah, of course. Of course

... it turns out, but it's, uh, it's quite small. I don't disclose technically, but, um, I think that solar is still... They have a, they also have one where it's like this thing could b- be, like, the face of it's just a solar cell, and, like, that is more realistic.

Or kinetic.

The kinetic, apparently, they, they... I'm not an expert in this, but they, they, they seem to think it wouldn't be enough. Uh, kinetic's quite small, I guess, on the capture.

Well, I mean, watchmakers have been powering with kinetic for a long time.

Yeah.

Anyway, we don't have to talk about that. Uh, y- I just wanted to get a sense of CES. Uh, would you do it again?

So-

I definitely would. I, I would.

Okay, you're just a fan of CES. As a business point of view, does it make sense? I happen to be in the conference business, right?

Yeah.

So I'm kind of just curious about that.

Yeah. So I would say as we did, so without the booth, and really, like, straightforward conversations that were already planned, three days, that's okay. I think it was okay. Uh, but if you need to invest for a booth that is not a good one-

Which is how much?

And, like, I think-

10 by 10 is 5,000-

But on top of that-

And then you go, like-

... you need to-

Financially

... 10 by 10 is, like, super small

Pre-fab the things.

Yeah.

Yeah.

And like, like some, some companies have, I think would probably be more in, like, the six-figure range to get. And I mean, I think that, yeah, it's very noisy. We heard this, that it's very, very noisy. Like, obviously, if you're, everything is being launched there, and, like, everything from cars to cell phones are, are being-

Yeah

... launched, so it's hard to stand out. But, like, I think going in with a plan of who you want to talk to, I feel like-

That was worth it

... worth it. We, we, we had a lot of really positive, um, media coverage from it, and we got the word out, and, like-

Yeah, it was very good

... that's, so I think we accomplished what we wanted to do.

Yeah.

Yeah.

I mean, there's some world in which my conference is kind of the CES of whatever AI becomes.

Yeah, I think that-

Don't do it in Vegas.

Don't do it in Vegas.

Yeah, don't do it in Vegas. Uh-

That's the only thing. I didn't really like Vegas.

S- SF is your carriage.

Your carriage.

Okay, that's great. Amazing. Those are my favorite ones.

But you, you cannot fit 90,000 people in SF. That's really the-

You need to do, like, multiple-

Yeah, yeah, yeah

... locations. So you could do Moscone and then have one in-

I mean, that's, that's what Salesforce, uh, conference is.

GDC is how many?

Dreamforce. Yeah, yeah.

That might be 50,000, right?

Yeah, yeah. Okay, form factor, right? Like, my way to introduce this idea was that I was at the launch in, in Solaris. What's the old name of it? Uh-

Newton.

Newton. Uh, of Tab, when Avi-

Yeah

... first launched it.

I remember.

He was like, "I've thought through every form factor. Pendant is the thing."

Mm-hmm.

And then we, and we got the pendant for this.

Yeah.

This original, the first one was just pendant, and I took it off and I forgot to put it back on. So you went through pendant, pin, bracelet now, and, you know, maybe there's AirPod, there's a sort of earphones in the future, but-

Mm-hmm

... what was your iterations through that?

Yeah. So we had, I believe now, three or four iterations, and one of the things that we learned is indeed that people don't like the pendant. In particular, woman, you don't want to have, like, anything here on the chest because maybe you have, like, other necklace or any other, other stuff.

Um-

You just ship a premium one that's gold.

Yeah, exactly.

That's what Apple does, right?

We're talking some, some, some fashion-

Some, yeah

... reached out to us.

Some, some big fashion. There is something there.

This, this is where it helps to have an Italian on the team.

Yeah.

Exactly. Exactly.

There is, like, some big Italian luxury. I can't say anything star . So yeah, bracelet actually came from the community because they were like, "Oh, I don't want to wear anything, like, as necklace or as a pendant." Like it's...

And also, like, the one that we had, I don't know if you remember, like, it was, like, circular. It was like this, and was, like, really bulky. Like, people didn't like it.

It's so ugly.

Yeah.

And also, I mean, I actually, I, I don't dislike. Like, we were running fast when we did that. Like, our, our thing was, like, we wanted to ship them as soon as possible, so we're not overthinking the form factor or the material.

We were just wanna be out. But after the community organically, basically all of them were like, "Well, why you don't just don't do the bracelet? Like, it's way better. I will just wear it, and that's it." So that's how we ended up with the bracelet, but it's still modular, so I still wanna play- Around the fact that it's modular and you can, you know, take it off and wear it as a clip, or in the future, maybe we will bring back the, the pendant.

Um, but I like the fact that there is some personalization. And right now we have two colors, yellow and black. Soon we will have other ones. So yeah, we can play a lot around that.

I think the form factor-- like, the goal is for it to be not super invasive, right? And something that's easy. So I think in the future, smaller, thinner, not like Apple-type o-obsession with thinness, but it does matter, like the, the size and weight.

And we would love to have more context because that will help. But to make it work, I think it really needs to have good power consumption, good battery life, and, you know, like with the Humane swapping the batteries.

I have one. I mean, I'm, I'm-- I think Humane is, like, pretty incredible, some of the engineering they did, but, like, it wasn't kinda geared towards solving the problem.

Mm-hmm.

It was just, um, it's too heavy. The swappable batteries is too much to man-

The heat

...like the heat-

Yeah

...the thermals. There's like too much to-

The light interface thing.

Yeah, like that-

That was like-

It was cool.

It's cool. It's cool, but it's like if-

But-

...if you have your hand out here, you wanna use your phone. Like, it's not really solving a problem 'cause you know how to use your phone. It's got a brilliant display. You have to kinda learn how to gesture this low-resolution laser.

Mm-hmm.

But the laser is cool, the fact they got it working in that thing, even though if it did overheat. But like, um, too heavy, too cumbersome, too complicated with the multiple batteries.

Mm-hmm.

So something that's power efficient, kind of thin, both in the physical sense and also in the edge compute kinda way so that it can be as unobtrusive as possible.

Yeah. Users really like-- Like, I like when they say, uh, "Yes, I like to wear it and forget about it."

Yeah.

Because I don't need to charge it every single day. On the other version, I believe we had like thirty-five hours or something, which was okay, but people, they just prefer the seven days battery life and-

Oh, this is seven days?

Yeah.

Oh, I've been charging ev-every three days.

I don't know. You, you can like keep it like-

But I don't know

...yeah.

Yeah, yeah.

It's like almost seven days.

The other thing that may-- that occurs to me, maybe there's an Apple Watch strap.

Ah.

Yeah.

So that I don't have to double watch. I-

Yeah, yeah.

Yeah, that's the other one that, yeah, I thought about it. Um, I, I saw as well the ones that, like, you can, like, put it, like, back on the phone. Like, you know-

Plog

...there are-- like there is a lot.

So yeah, there's a competitor called Plog.

Yeah.

It's not really a competitor. They only transcribe, right?

Yeah, they only transcribe.

But they're very good at it.

Yeah.

They're-- No, they're great. Their hardware is, is really good too. But-

And they just launched the pin too.

Yeah. I think that, um, the MagSafe kind of form factor has a lot of advantages, but some disadvantages.

Yeah.

Um, you can definitely put a, a very huge battery on that, you know? And so, like, the battery life's not-- the power consumption's not so much of a concern. But, you know, downside, the phone's, like, in your pocket and...

So I think that, you know, form factors will continue to evolve, but-- and you know, more sensors, less obtrusive and-

Yeah, we'll have a new version

...be easier, easier to use.

Soon.

Okay. Looking forward to that. Um, yeah, I mean, we'll-- whenever we launch this, we'll try to show whatever, but I'm sure you're gonna keep iterating.

Mm-hmm.

Last thing on hardware, and then we'll, we'll go onto the software side 'cause I think that's where you guys are also really, really strong. Vision. You wanted to talk about why no vision.

Yeah, I think it comes down to, like, when you're, um, when you're a startup, especially in hardware, you're, you're just-- you work within the constraints, right? And so, like, vision is super useful and super interesting, and what we actually started with.

There's two issues with vision that make it, like, not the place we decided to start. One is m-power consumption. So you know, you kinda have to trade off your power budget. Like, capturing, even at a low frame rate and transmitting.

The radio is actually the thing that takes up the majority of the power, so you would really have to have quite a, like a unacceptably, like, large and heavy battery to do it continuously all day. We have, I think, novel kinda alternative ways that, that might allow us to do that, and, and we have some prototypes.

The other issue is form factor. So, like, even with, like, a wide field of view, if you're wearing something on your chest, it's, it's going-- You, you know, you-- Obviously, the wrist is not really that much of an option.

And if you're wearing it on your chest, it's, it's often going to probably be not capturing, like, the field of view of what's interesting to you. So that leaves you kinda with your head and face, and then anything that goes on, on the face has to look cool.

Like, I don't know if you remember the spectacles. That was kinda like the first-

Yeah, yeah. I have a pair

...spectacle. Yeah, but they kinda-- They didn't-- They were not very successful, and I think one of the reasons is they were-- they're so weird looking.

Yeah, the camera was so big on the side.

Yeah. And if you look at the Meta Ray-Bans, where they're way more successful, they, they look almost indistinguishable from a Ray-Ban. And they invested a lot into that, and they, they have a partnership with Qualcomm to develop custom silicon.

They have a, a stake in, uh, Luxottica now.

Luxottica.

So, like they-- Coming from all the angles. Like, to make glasses, I think, like, you know... I don't know if you know Brilliant Labs. They're a cool company.

Heard of them.

They make frames, which is kinda like a cool hackable glasses and, and-

Hmm

...and, like, they're really good. Like, on hardware, they're really good. But even if you look at the frames, which I would say is, like, the most advanced kinda startup, I mean, there was one that launched at CES, but it's not shipping yet.

Like, one that you can buy now, it's still not something you'd wear every day, and the battery life is super short. So I think just the challenge of doing vision right, like off the bat, like, would require quite a bit more resources.

And so, like, audio is such a good entry point, and it's also the privacy around audio. If you, if you had images, that's like another huge challenge-

Mm. Yeah, yeah

...to overcome. So I think that ideally, the personal AI would have, you know, all the senses and, um, you know, we'll, we'll get there.

Yeah.

Okay, one, one last hardware thing, 'cause I have to ask this because, um, uh, then, then we'll move to the software. Did-- Were, were either of you electrical engineering?

No, I'm CS. Um, and so-

CS. Okay

...uh, I have a, um- I've taken some EE courses, but I, I had done, prior to working on, on the hardware here, like, I had done a little bit of, like, embedded systems. Like, very little firmware, but we have luckily on the team-

Yeah

... somebody with deep experience. Um-

Yeah. I'm just like, you know, like, you have to become hardware people.

Yeah, yeah. I mean, I, I learned how-

You need to, you need to learn

... you have to worry about supply chain-

Yeah

... power-

I think this is like radio

Yeah. There are so many things to learn.

The, the, the... I would tell this about hardware, like, and I, I know it's been said before, but-

Hardware's hard.

Like, building a prototype and, like, learning how the electronics work and learning about firmware and, and developing, this is, like, I think, fun for a lot of engineers, and it's, it's all totally, like, achievable, especially now, like, with, with the tools we have.

Like-

Oh, I mean it's-

... stuff you might have been intimidated about, like, how do I, like, write this firmware now? With Sonnet, like, you can, you can get going and actually see results quickly. But I think going from prototype to actually making something manufactured-

Yeah

... is a m- enormous jump, and it's not all about technology. The supply chain, the procurement, the regulations, the cost, the tooling. The thing about software that I'm used to is it's funny that you can make changes all along the way and ship it.

But, like, when you have to buy tooling for an enclosure, that it's expensive.

You, you buy your own tooling?

You have to.

Don't you just subcontract out to someone in China?

Oh, no. Do we make the tooling? No. No.

Oh, okay.

You have to have CNC and, like, a bunch of machines. Like, nobody makes their own tooling. But, like, you have to design.

Yeah.

You design.

This design, and you submit it-

You ship the design

... and then they go four to six weeks later.

Yeah.

And then if there's a problem with it-

Right

... well, then you're not-

You're fucked

... you're not making any-

Yeah

... any of your enclosures.

What resources-

And so you have to really plan ahead and, like-

I, I just wanna leave tips for other hardware founders. Like, what resources or websites are mo- uh, most helpful in your sort of manufacturing journey?

You know, I think it's different depending on... Like, the, its hardware, it's so specialized in different ways.

I will say that, for example, like, to choose a manufacturer company, speak with other founders and ask-

Yeah. Yeah

... ask maybe us. Like we can give-

Background checks

... you, like, some-

Yeah

... you know, some tips of who is good and who is not, or, like, who specialize in something versus somebody else.

Yeah. Like, some people are-

Yeah

... good in plastics, some people are good on-

Some others are good on something else

... the PCB.

I think, like, for us, it really helped at the beginning to speak with others and understand, okay, like, who is around. I worked in Shenzhen. I lived almost two years in China. I have an idea about, like, different hardware manufacturer and all of that.

Soon I will go back to Shenzhen to check out. Uh, so I think it's good also to go in place and check and choose the right people.

Yeah, you have to, like, once you, when, if you-

Yeah.

So we did some stuff domestically, and, like, if you have that ability, the reason I say ability is it's very expensive. But, like, to build out some proof of concepts and do field testing before you take it to a manufacturer.

Despite what people say, there's really good domestic manufacturing for small quantities at extremely high prices. So we got our first PCB and the assembly done in LA, so there's a lot of good 'cause of the defense industry that can do quick churn.

So it's like, we need this board, we need to find out if it's working, we have this deadline, we want to start. But you need to, to go through this and, like, if you wanna have it done and fabricated in a week, they can do it-

Yeah

... for a price. But I think, you know, everybody's kind of trending even for prototyping now, moving that offshore. Because in China you can do prototyping and get it within almost the same timeline. But the, the thing is with, um, manufacturing, like, it really helps to go there-

Yeah. For sure

... and kind of establish the relationship.

Yeah. My first company was a hardware company, and we did our PCBs in China, and took a long time. Now things are better, but this was, yeah, I don't know, 10 years ago, something like that.

Yeah, I think that, like, the... And I've heard this too. We didn't run into this problem, but, like, you know, if it's something where you don't have the relationship, they don't see you, they don't know you, you know, you might get subcontracted out or, like-

Yeah, yeah, exactly

... they're not, they're not paying attention. But, like, if you're, you know, you have the relationship and a priority, like, yeah, it's, uh, it's really good. We, we ended up doing the, um, fabrication assembly in Taiwan for various reasons.

Uh, but-

Yeah, and I think it really helped the fact that you went there at some point.

Yeah, yeah, yeah. And we're really happy with the process and... But, I mean, the whole process of just-

Choosing the right people

... choosing the right pe- but also, um, just, uh, sourcing the bill of materials and, and all of that stuff. Like, I guess, like, if you have time, it's not that bad, but if you're trying to, like, really push the speed, it, that, it's incredibly stressful.

Okay, we gotta move to the software.

Um-

Yeah, yeah, yeah. So the hardware, maybe it's hard for people to understand- ... but what software people can understand is that running transcription and summarization, all these things in real time, every day, for 24 hours a day, it's not easy.

So you mentioned 200,000 tokens per day.

Yeah.

How do you make it basically free to run all of this for the consumer?

Well, I think that the pipeline and the inference, like, people think about, oh, these tokens, but as you know, the price of tokens is, like, dramatically dropping. You guys probably have some chart somewhere, uh, that you've, you've posted.

We do. Yep.

And, like, if you see that trend, in, in, like, 250,000 input tokens is not really that much, right? Like, the output to-

And you do several layers. You do live.

Yeah, yeah. So the, the speech to text is, like, the most challenging part actually, because, you know, it requires, like, real-time processing and then, like, later processing with a larger model. And one thing that is fairly obvious is that, like, you don't need to transcribe things that don't have any voice in it, right?

So good voice activity is key, right? 'Cause, like, the majority of most people's day is not spent with voice activity, right?

Yeah.

So that is, uh, the first step to cutting down the amount of compute you have to do. And voice activity is a fairly cheap thing to do. Very, very cheap thing to do. The models that need to summarize, you don't need a Sonnet-level kind of model.

To summarize, you do need a Sonnet level model to like execute things like the agent and, um, we will be having a subscription for like features like that because it's-

Yeah

... you know, although now with the R1, like we'll see, uh, we haven't evaluated it.

A DeepSeek? Yeah.

Yeah.

I mean-

I mean, not, not that one in particular, but like, you know, they're already there that can, can kind of perform at that level. I was like, "I was gonna say in six months," but like, yeah. So self-hosted models help in, in the things where you can.

So it's-- you are self-hosting models?

Yes.

You are fine-tuning your own ASR?

Yes. Um, I will say that I see in the future that everything's trending down, although like I think there might be an intermediary step with things to become expensive, which is like we're really interested because like the pipeline is very tedious and like a lot of tuning, right?

Which is, is, is brutal because it's just a, a lot of trial and error. Whereas like, wow, wouldn't it be nice if an end-to-end model could just do all of this and learn it?

Mm-hmm.

If we could do transcription with like an LLM, there's so many advantages to that, but it's going to be a larger model and hence like more compute. You know, we're optimistic maybe we could distill something down and like we kind of more than focus on reducing the cost of the existing pipeline or trying to the next generation-- 'cause it's very clear that like all ASR, all speech to text is gonna be pretty obsolete pretty soon.

So like investing into that is probably kind of a dead end 'cause it's just gonna be obsolete.

It's interesting. Like, uh, I think when I initially invested in Tab, uh, this is-- this shows you how wrong I was. I was like, "Oh, this is, um, a, uh, sort of razor blades-- blade-- razors and blades model, where you sell a cheap hardware and you make up a subscription, like a monthly subscription."

And now I just checked, Friend is a one-time sale, $99. Limitless, one-time sale, $99. These guys, one-time sale, $49 and, uh, inference is free. It-- what?

Well, would you-

It's crazy.

I think when you probably invested like how much was a, a million input tokens at that time and what is it now?

It's a fascinating business and like, uh, you know, there's a lot to dig into there, but just getting that perspective out there is I think it's not something that people think about a lot, and you obviously have thought a lot about.

What about memory? I think this is something we go back and forth on about memory as in you're just memorizing facts and then understanding implicit preference and adjusting facts that you think about a person. Any learnings from that?

I know there's a lot of open source frameworks now that do it.

Yeah.

Did you build all of your own infrastructure internally?

Yeah, we did. I mean, I evaluated-- used a lot in other projects. I think that there's a few different tasks or, or things that revolve around memory. Like one is like retrieval obviously, and like when you need to find, like even if you have a large corpus of how do you find and so, um, like I think existing kind of RAG pipelines also will probably be obsoleted.

The frameworks I, I have not found one-- like there's no general way to do RAG that works. Like it's really highly dependent on the data. So like if you're gonna be customizing something that much, it's just you get kind of more bang for the buck from designing it all yourself.

You know, a lot of those frameworks are great for getting, getting going quickly. Um, but I think it's really interesting memory when you're trying to do-- for a person 'cause memories decay, right? Like I'm going to London, you know, then I come back.

I'm not going to London anymore. What we've learned is like doing the traditional like embedding and, and RAG is suboptimal. We, we kind of built our own using small models to do really p-- massively parallel retrieval, which I think is gonna be maybe more common in the future.

And then like how to represent a person. We still require some human in the loop, and I mean, this is an ongoing project and, you know, we're, we're, we're learning every day on like how do you correct the, the model when it gets something wrong about you.

Right now we have like things that are like super confirmed that are like ground truth about you because the human accepted it. But ideally, like that step wouldn't be necessary. And then we have things that are fuzzier and like the more stuff that we know is true, the, the more accurate we are when we're trying to decide is this fuzzy stuff 'cause it's probably like if you have the context, it's probably not true.

So I think it's one of the most core challenges is how to handle both retrieval and then modeling and like, especially when you're dealing with noisy source data 'cause like even if-

Yeah

... in an ideal world, even if you just had perfect transcription and you're going off that, that's still not enough information, right? And even if you had visual, it's still not enough. Like there's still gonna be some misunderstandings and so how to not let that damage the value of it and be like recoverable and correctable.

Yeah, one way I think about it is I usually like to order the same thing from the same restaurant if I like it, but I'm, I'm not saying that out loud. And it's kinda like, are these type of behaviors-- like when you ask about a favorite restaurant, I would just want it to give me restaurants that I've already been to that I like.

Or like if I'm like, "Hey, just order something from this place," it should just reorder the same thing because it knows that I like-

Yeah

... to redo the same thing. But I feel like today most agent memory things that I see people publish, it's like, you know, just write down the data-

Yeah.

Yeah.

Yeah, yeah.

Yeah, I mean, I think that's why the reasoning like in, in our case, like giving it time to consider all of the sources it has. So like look at the emails, see like the receipts, and then look at the conversations to see like what I've mentioned, and then be able to then take enough time to search through all the context and connect the dots-

Mm-hmm

... is I think really important. And like I don't know, like some of the agent memory stuff it's, it's like the key value with RAG on top like and the results there are just not ... complete enough when you have, like, growing corpus and, like, managing decay and, and hallucinations that might be in the source material and

So this is where people usually bring in knowledge graphs.

Yes.

Mm-hmm.

And do you do it?

We don't extensively use knowledge graphs. It's something, you know... We, we didn't talk also about the kinda potential future social aspects-

Yeah, I want us to speak about it

... but, like, um, but the, the problem with knowledge graphs that we found is like... And I don't know, um, if you can tell me what, what your experience has been, but they're great for representing the data, but then, like, using it at inference time is kinda challenging, like-

For speed or what other issues?

Just, like, the LLM understanding, like, the, the graph-

Yeah

... the input.

Yeah, it's not in the training data for sure.

I think that the graph is the right kinda way to store the data, but, like, then you need to have the right retrieval and then just kind of formatting in a way that, like, doesn't just overwhelm or confuse-

Okay

... what you're trying to do.

Should we ask about social?

Yeah.

Yeah, no, I, I thought you were gonna go into it.

Yeah.

Yeah, what's the-

Yeah.

Like, not directly related-

We did some experimentation-

Yeah

... that was quite-

Not directly related to, like-

... interesting

... graph retrieval or graph knowledge bases-

Yeah

... but, like, the idea that having, like, your personal context, but then, like, other people can query it, you know, it can divulge some things that you would have full control over. Then Maria and I are trying to negotiate, like, where we're going to dinner, like, the, the-

We, we exactly did this experiment

... there can be an exchange. Yeah.

Yeah.

There can be an exchange between the, the agents and like-

So how-

Oh

... like, my agent can speak with Ethan's agent, both of them, they know our location-

Okay

... what we like, where we went in the past.

Yeah.

And even, you know, if we have our calendar integrated, they know when we're free, so they can interact to each other and have a conversation and decide a place to go for us.

Wow.

And we did that, and it was for me really cool because they suggested to us a nice French restaurant that we went at the end.

That you never been to?

That we never been to.

Okay.

Uh, but both of us, they said that we like French food, both of us. We were in Pacific Heights. And yeah, this was really trivial.

Yeah, it's a trivial, like, toy use case, but I guess, like, in terms of you've been using it for a while, like, if I wanted to buy you a gift-

Oh my God, you bunch me... You, you bought me a bunch of candles now that I think about

Yeah, yeah.

Yeah.

There's another use case. I was like-

Based on that.

Yeah. And when we were testing the agent-

Yeah, yeah

... like, a bunch of candles from Amazon showed up at her door.

Yeah, because I really, I love candles-

Yeah, yeah

... but I didn't expect 20. Um, but yeah.

Yeah, there's a lot of experimenting. But, like, how to manage that, where it's like what's okay for your-

Yeah

... your Bee to divulge to who.

Yeah, you need-

Like, should you get a authorization request-

Yeah

... every time?

Oh, oh, scopes-

Yeah, yeah, yeah

... for personal context.

Yeah, yeah, yeah.

Mm-hmm.

So, like, you know, you would have to... Human would have to sign off on it, but, um-

Yeah

... I think then, like, then I wouldn't have to guess.

Mm.

I could just...

Yeah, yeah. You know, there's this culture that, like, is very alien to everyone else outside, outside of SF and outside the Gen Z bubble in SF, which is sharing, location sharing.

Yeah.

Yeah.

Like, I can tell my close friends where they are exactly right now in the city.

Yeah.

Yeah.

And it's opt in, and, like, it's mu- you know, and, like, it's normal, and, like, it freaks out everyone who's not here.

Yeah. Yeah.

And so maybe we can share preference, like-

Yeah

... who we'd like what.

Absolutely.

Yeah, what we want.

I, I really believe in it, for sure.

Yeah.

We will.

Or even, like, small updates about your day, like.

My parents would love that-

Yeah

... 'cause I don't, I don't do that.

Yeah, yeah. So, like, now, now there's no friction. It can just be more or less automatic and-

Yeah, yeah. Uh, dating?

I was trained always to avoid dating-

Really?

... uh, as a startup founder.

Yeah, you kind of hate that. Yeah.

Uh, but-

Everyone hates it?

We thought about it. Like, uh, sometimes some people, they ask to us because it's like, "Oh, you know so much about me. Like, can you measure compatibility with somebody else or something like that?" Yeah, probably there is a future.

Maybe somebody should build that. I think on our end we were like, "No, this is-

Okay

... we, we don't wanna risk."

Like, I will build on your API.

My sister's actually a personality psychology professor, and she studies personality. And we were at Thanksgiving, b- 'cause my parents were one, and, and I was like, "Ask it, like, give me my, um, big five"-

Yeah

... which is, like, the personality type, and it's like-

Does it know my big five?

Just, just-

Yeah, probably

... just ask it to consider everything and, and give your big five. And my sister said it was pretty... I, I didn't agree with it 'cause it said I was disagreeable, but she-

I agree with that.

... she seemed to think I was agreeable. Um, and so-

You disagree that you're disagreeable?

Yeah, yeah.

What other proof do you need then?

Yeah.

I think I'm very agreeable. Um. Um, but I think that we did get some users are like, "Oh, if, if, like, we're a couple."

Yeah, we had, like, couples actually. They bought the product together.

Yeah.

They both-

Yeah

... like couple, they bought our hardware, so there is something there. Another test is, like, the Myers-Briggs. I know that you don't like that one. Um, but-

No, no, Ocean is cooler than Myers-Briggs.

Yeah.

Yeah.

Everyone, stop using my MBTI. Use Mys- use Ocean.

Yeah, yeah. For me, like, it was on point, like, every time, like, it...

Awesome. Anything else that we didn't cover? Any cool underrated things?

Go to bee.computer, 49.99, and you buy the device. That's the... No.

That's the call to action.

That's the-

And you're hiring?

We are hiring, um, for sure.

AI engineers.

AI engineers.

Nice. Yeah.

What is an AI engineer?

Yeah.

But you- did you study?

So- so- somebody who, who's scrappy and willing to-

Work with us

... yeah. I, I think- I think that you coined the term, right? So you can tell us.

Somebody-

I mean, I, I-

... that can adapt-

It's really, like-

... that has resistance

... people have different perspectives and what-

Yeah

... is useful for you is different from what is useful for me.

Mm-hmm.

Yeah, so anyway, it's, it's all useful.

I mean, I think that always-on AI is really gonna explode, and it's gonna be a lot from both a lot of startups but incumbents, and there's gonna be all kinds of new things that we're gonna learn about how it's gonna change all of our lives.

I think that's the thing I'm most certain about, so...

And be an AI. Well, thanks very much.

Thank you, guys so much.

Yeah, this was a pleasure.

Thank you.

Uh, yeah. We'll see you launch whenever that launch is happening.

Yeah.

Yeah. Thanks.

Thank you.