0:00
Can you tell me about the Access podcast?
Search this video
0:02
Yeah, Access is a tech industry [music] podcast hosted by Alex Heath and Ellis Hamburger.
0:06
They interview big players and emerging founders, kind of the inside baseball of Silicon Valley.
0:12
Okay, this product is dope.
0:14
This week on the show, we're sitting down with Mina Thami, the co-founder of Sandbar.
0:18
It's a startup building what might be the most interesting AI wearable you've never heard of.
0:22
A smart ring that doesn't just capture your thoughts, but talks back to you in your own voice.
0:29
We get a live demo of the hardware, we dig into Mina's background and neural interfaces at MIT and Meta, and then we talk about why the future of AI might literally be on your finger.
0:41
Plus, Ellis and I get into what's happening inside OpenAI right now and why the race between ChatGPT and Claude is about to get a lot more interesting. Welcome to Access.
0:55
>> [music] >> Welcome to Access.
0:56
I am Ellis Hamburger, here with my co-host Alex Heath.
0:59
What's going on, my friend? I'm good, man.
1:01
You were actually in Mexico right now. Technically.
1:05
[music] Back to the future.
1:07
Using a rift in the space-time continuum.
1:08
By the time this episode comes out, you will have just come back from hopefully a very relaxing beach vacation. I sure hope so.
1:17
Because >> [laughter] >> I think ever since I started meaning a few years ago and going through this renovation that went a lot over budget and etc. etc.
1:27
kids, I haven't really had an actual vacation in like years aside from like family commitments.
1:33
So yeah, I'm hopeful that this Cabo resort will literally do everything for me and everything for us.
1:41
We going to have some recreational drugs?
1:43
What are we going to be doing down there?
1:44
Recreational Miami Vice cocktails. Yes.
1:50
Um I mean, do you unplug when you're actually on vacation? It's hard, man.
1:51
I it takes me I found, especially since starting, you know, a company, it takes me almost 2 days to fully let myself relax.
2:04
So when you do the like short weekend or long weekend, you know, getaway, which I've done in Cabo several times, I'm usually relaxed by, you know, the last half of the day before I leave.
2:16
>> [laughter] >> Where is the phone located is the question, cuz I feel like I went on this nature retreat during college for like 7 weeks that was just the absolute best and we had no technology or electricity and it probably took like four or five days to stop feeling those phantom buzzes on your leg from the phone vibration.
2:35
And even longer to like stop reaching for my phone every time I had a question that I wanted to Google or something like that.
2:45
You just had to be like, "Oh well, I guess I'll show you this song in 8 weeks." Yeah.
2:52
Uh the phone is definitely near me.
2:52
I'm usually cuz I I I catch up on podcasts when I'm on vacation.
2:56
I like to listen for in a more like active way where I'm actually like thinking and then it becomes work cuz I'm thinking about, "Oh, I like this idea.
3:04
I like the way they did this. We should do this."
3:05
I think I was literally texting you like pod ideas probably stoned at the pool in Cabo last time I was there.
3:13
But that's what all good founders say.
3:15
I mean, this is the same as with Josh Miller at the browser company or Evan Spiegel at Snapchat.
3:22
They go on vacation and then they think of what they want to do for their 2.
3:23
0 or this or that and they're like, "Why don't I do this more often?"
3:27
And it's like, yeah, I mean, being away from everything gives you the free time to actually let your subconscious do its thing.
3:34
And I mean, I guess that's why you want the Stream Ring so bad. Exactly.
3:37
We've got Mina, the founder of Sandbar and this very cool ring, which I think we've talked about on the show before.
3:45
Maybe we can get to that.
3:47
Um getting ready to come out soon and we both know Mina, have hung out with him.
3:53
He came and showed me the original prototype, or not the original, but one of the earlier prototypes actually right by your house um in LA at a cafe about a year ago and it's cool to see the progression.
4:05
They've raised a lot of money, they've got a great team and really unique, interesting take on AI on your on your finger.
4:11
I don't know, he's a very thoughtful dude, you know. Yeah. Yeah.
4:16
I love a good love a good design founder.
4:18
Clearly one of those founders who's just truly thinking deeply about the impact of their product on society and I mean, I meet with founders every single day and, you know, a lot of them are just kind of here for fun.
4:34
>> [laughter] >> Want to build stuff.
4:37
But yeah, they seem to have the full package across hardware.
4:39
He talks on the show about how they've got folks from Apple, from Fitbit, from Peloton and then on the software side as well. It's very cool to watch.
4:49
I mean, it's been an absolute joy as a gadget nerd to see kind of the rise again of hardware startups, which we I feel like we we didn't see for quite a while.
4:59
But yeah, they're they're back in full force.
5:01
There's a lot of noise out there, but Mina's a pretty quiet dude and they've not made huge splashes, but I'm very actually bullish on this form factor as we talk about.
5:12
It's hard it's been I've been thinking a lot about it and how I want it since I first tried it.
5:15
Um but yeah, man, before we get to that, I guess the only thing I've got is by the time this comes out, it will have been about a week since this, but yesterday I spent a chunk of the day at OpenAI with a lot of their leadership, especially on the Codex, their you know, Claude Code competitor side.
5:32
And then went to a pretty cool dinner um with the Codex team and Greg Brockman, um the co-founder of OpenAI at their original office, which was cool in in the library where they did a lot of earlier research work.
5:48
Uh sushi dinner there in the library, which was cool.
5:50
Uh the dinner was off the record, but I had a bunch of interesting on-the-record conversations with them and including Peter Steinberger, the creator of Open Claude, the Claude father, who is about 3 weeks in to being at OpenAI.
6:02
So he told me kind of what he's working on and what he's thinking about his first interview.
6:08
Is that your current crush in the tech industry if you had to choose?
6:13
Yeah, Peter's an awesome dude.
6:13
He's he's very he's very genuine.
6:14
He's just a old-school nerd who's in it for nerdy reasons and it's very clear when you talk to him. >> Uh he's huge. He's jacked.
6:24
Yeah, >> [laughter] >> I felt scared to do that.
6:26
Um but uh it was cool, man.
6:30
It was his first interview since officially joining joining and that that's going to be out on Sources and then also by the time this comes out, my broader piece about what what is really happening is like Codex is taking over that entire product suite, right?
6:43
Like they killed Sora, you know, the week before this episode comes out and that's part of them focusing all their efforts and GPUs most importantly on building this super app to counter what Anthropic has done with Claude and Claude Co-work, which has really abstracted away coding for normies where you are coding, you're using AI to execute code and crawl databases and all this stuff, but you don't need to know that.
7:10
You can look at it, but you don't and that's a very powerful thing.
7:13
I mean, I use it every day all the time. I know you use it.
7:15
And what chat is going to become is that essentially and I think that's actually the future of all these assistants that we're going to be talking about in probably more six-ish months time frame is when this is going to really start catching on beyond the, you know, early adopter crowd that's listening to this show.
7:33
But um it's going to be really interesting when you start building these coding harnesses in with the consumer chatbot experience and that is very much underway at OpenAI as a response to what Anthropic is doing.
7:47
Well, I mean, you know the personalities involved here.
7:49
I mean, is this wanting to get back on the hype cycle?
7:54
Is this seeing the financial opportunity with Claude integration at companies and how much bigger that is than something like Sora?
8:01
Or is this an actual change of heart?
8:03
Because I mean, as we talk about with Mina, you know, your products kind of and your product strategy represent your beliefs around what you're trying to do.
8:11
And I mean, whether, you know, when we talked with Fiji, whether it was about Sora and creativity or pulse and personal utility and getting these different, you know, news sources fed to you every day.
8:20
I mean, has there been a change of heart there about what the company is about?
8:24
I think they regret not being so good at coding so early as they could have been.
8:33
Anthropic focusing on coding has really given it this take-off moment that it's in right now in the in the the UI of Claude Co-work making it easy for people.
8:41
I would I think they wish they got into that first.
8:43
I don't think it's a change of heart.
8:45
I think it's seeing, "Oh, this is taking off.
8:46
The models are also now capable of doing agentic things reliably and or at least most of the time reliably.
8:55
And so the moment is now and I think they're looking at the growth of Claude Code and going, you know, yikes. We need to combat that.
9:04
And I mean, these companies do believe that coding is a way to get to AGI, which is very that's definitely not a pivot.
9:10
They've been trying to get to that, but um I I do think OpenAI is is realizing that the money and the gains and the hype and the good vibes are in this coding productivity world right now and not everything else, not adult mode, which they told me exclusively, right, that they're shelving. Sora, which they killed.
9:28
They're kind of getting rid of what they're calling side quests to focus on this super app, which Brockman, the co-founder, is kind of leading with Fiji.
9:38
And it means that Codex is going to be the the harness, literally, but also just kind of enveloping ChatGPT in the same way that Claude Code is taking over Claude. Yeah, I mean, we'll see.
9:51
You know, I think if you're following the trends and the hype and the crowd, I think that is arguably how you end up in the kind of situation that Meta's in, where they go on these shopping sprees every 2 years.
10:03
Buy up all the latest trends or whatever. I don't think it's that.
10:07
Cuz everybody was saying everybody was saying that Sora is Sora was going to be the next big thing and everybody's going to be making AI videos like with full conviction.
10:14
And then it just kind of didn't happen.
10:16
I mean, Sora Fiji talked to us about that.
10:20
Like it it was an early thing and it wasn't growing, I think the way that maybe had hoped it would continue to grow after it had that initial pop and was at the top of the App Store.
10:30
Meanwhile, people are using the out of these AI coding tools and Cloud Co-work, right?
10:34
And like you can do some of that stuff in Codex, but there's an awareness problem where people don't know that Codex is capable of Cloud, you know, Co-work's abilities.
10:44
And I think they've got to build that into the chat interface, which is, you know, reaching almost a billion people. Which is a huge move.
10:50
And so basically, I just think big picture like we're not going to be thinking of these chatbots like kind of the basic, you know, gets put some text in, get some text back or maybe some visual chatbots.
11:00
We're going to be thinking of them in 6 to 9 months as like agent orchestrators.
11:06
And that's going to introduce this capability to a lot more people very quickly. Yeah.
11:13
And where do you feel like it's going to net out that kind of distinction between the actual apps that you might need?
11:19
Like whether it's with that company Dreamer that was acquired by Meta or something like Wabi or even what some of the others like Replit are creating with your own applications.
11:27
Like do you think those are going to be a part of this same linear, you know, model or are they going to be separate?
11:34
Cuz it's a lot of the same features, right?
11:36
Some of them just have interfaces and some don't in terms of the orchestration.
11:39
I think a lot of the vibe coding tools, Replit or whatever, are already connected via MCP to, you know, the big assistants.
11:47
to, you know, the big assistants. I think the assistants, when I say that, I mean ChatGPT and Claude mainly, but Gemini as well, are going to be these orchestration layers that have all your contacts, your memory, hooks into your data
12:02
and your tools, which are Figma, which is like really opened up via MCP to Claude or um you know, Replit, which, you know, I was saying with the granola MCP, um you can literally like just have, you know, we were talking about that last week with Sam, right? You can
12:16
You can have Replit spin up something, you know, via Claude as your granola is is coming or whatever.
12:23
Like I think we're going to start thinking about apps really differently uh in the next few months.
12:27
I think it's going to happen faster than people think um based on what the labs are saying.
12:32
I mean, it's it's it's just interesting kind of fuzzy because I mean, Eugenia from Wabi has, you know, described a lot of these mini apps as kind of like containers for prompts, uh for things that just often are better as a user interface.
12:47
Like for example, a photo gallery of AI generations is great as a little app or user interface.
12:53
Um and I think if you look at Dreamer as far as I could tell under the hood, it's a lot of connections being made between all these different services that all of themselves have APIs and MCPs.
13:05
And that's something that you could theoretically just ask Claude in a year to do. But I don't know.
13:12
When you try to I mean, how important were the interfaces for those things?
13:16
Or was it more just about templated orchestrations?
13:17
Yeah, Dreamer was a really cool platform sold to Meta in a month of being in private beta.
13:26
I got added about a week and a half before it sold.
13:28
Uh little inside baseball for how quick this Meta deal came together is the CEO was asking to be on this pod like a few days before the deal was announced.
13:38
So uh I don't think uh they knew this was coming.
13:43
Uh it's a great team, a lot of early Android, Stripe, Meta people and they were building agents for normies.
13:49
They were building a marketplace uh as well, where you could remix other people's agents and uh get rev share from how they use it, which I thought was really cool.
14:01
Uh and it was a an agent open Claude-ish type thing that like uh you know, your parent couldn't break, which I thought was cool.
14:08
And the way they had all these interesting data sources that I haven't seen in other chatbots, live game data from, you know, sports leagues, stuff like that.
14:17
Uh was was was a really unique approach and the interface was really simple.
14:21
And clearly Meta thinks that that's something they need to kind of build into their platform, uh whether it's for their businesses that advertise or or end users. So yeah, man.
14:33
It would've been cool to see Dreamer keep developing and, you know, have have David the CEO on um but uh they went and got those Zuck bucks. Good for them. That's funny.
14:42
That happened to me as well.
14:44
A million years ago, I was all set to >> No, I was all set to do a big, I forget if it was an exclusive or what with the creators of Vine. Mhm.
14:54
And then when I was at The Verge and then they completely ghosted me, which like pretty much never ever happened and which I'm maybe even still butt hurt about.
15:02
And I was like, where did they go?
15:04
This was really cool, you know?
15:06
And then a few weeks later, uh yeah, the announcement that that Twitter acquired them and I was like, oh. That's why.
15:13
Yeah, this happens every now and then where a founder goes silent.
15:17
Yeah, and you're like, that's why.
15:19
>> [laughter] >> Uh yeah, that's all that's on my mind right now.
15:23
Uh I hope you are enjoying your vacation this week.
15:25
I'm going to manifest it from today's recording to where I will be in a week.
15:31
You're going to be so relaxed when you come back and yeah, man.
15:33
I guess with that, we should probably kick it to Mina. Cool.
15:40
>> [music] >> Mina, what's going on, boss? Good to see you. How are you? Good to see you.
15:44
Are you Is this your headquarters you're coming to us from? It is indeed. Sunny Flatiron.
15:50
Got You got some nice lighting in there, some skylight?
15:52
We we're in a basement for about a year and so our single requirement for our new office was as much sunlight as humanly possible.
16:00
How does it feel to be a certified ring guy now?
16:05
You're one of those designing the future of rings.
16:07
Everybody's got to take I I wore I wore rings a bit actually before Sandbar, but I have stopped effectively wearing anything except for for for our prototype by this point.
16:18
I think I I may have just gotten tired of wearing things.
16:20
Well, I find it really interesting.
16:22
Whenever I meet someone, I try to ask why they're wearing the jewelry that they're wearing and there's usually very deep and personal reasons.
16:30
Like whether it's a bracelet, a necklace, a ring, it'll have been a gift from a loved one or it'll be a reminder for a period of life or it'll be some, you know, strange journey they went on where they wandered into a place and, you know, picked something off a shelf.
16:45
Um it's very rare that I meet someone and there isn't some very deep reason for what they're wearing.
16:49
So this is your version of the TikTokers who go up and go, "You look so confident."
16:53
Why are you But you go up to them and say, "Why are you wearing that necklace?"
16:58
That and ideally no camera.
17:01
It's certainly not a tough sell to us ring bros in LA.
17:05
Everybody's got a signet ring or some type of woo-woo uh spiritual jewelry. I'm naked. No bling?
17:16
I mean, I got a wedding ring, you know. >> Okay.
17:18
But uh you know, I'm not taking that thing off, but I'm I'm going naked otherwise on the fingers.
17:24
Yeah, that's that's a that's a very meaningful ring to be wearing.
17:27
Until I get your ring, which is soon, right? Yes.
17:30
Yes, we'll be shipping in the summer.
17:34
And we're actually going to be racing towards beta sooner than that.
17:37
We're aiming for beginning beta in April, where we'll begin distributing more prototypes to more folks to to to try out.
17:46
Yeah, so like walk us through.
17:49
What's it been like to actually wear it every day?
17:51
How long have you been wearing it every day now? Maybe around 2 years.
17:56
Um it depends on we've had a bunch of generations.
18:00
So I brought with me uh some of the older prototypes. This was the first one.
18:06
Um so uh you can actually see it's For our audio listeners, he looks like he's holding something that like the optometrist would put in front of your eyes maybe to like test your eyesight is what this looks like. Poetry.
18:16
Or I would say like an ankle bracelet when you're uh on house arrest or something like that, but for your finger. Very stylish.
18:25
Yeah, I was I was wearing it a lot on on the High Line um on the west side of of Manhattan, where like people will go out in their full regalia and, you know, this was meant to be a bit of a fashion statement, but when we've had this prototype and we were trying things out the use cases looked very different.
18:42
We had it hooked up to search APIs, we had it hooked up to navigation and we were mostly using it as kind of an urban exploration product.
18:51
Um and it was only when we began exploring conversation and notes in the moment that we felt the use case begin to click together.
19:01
So after a few more iterations, we started to give this out actually in what we called alpha to folks to test.
19:08
So this version also had a gap on the bottom and eventually we closed it and got to this version.
19:16
And that's where we saw people beginning to use it in different ways.
19:21
Some people use it as a personal CRM where like after meetings, they'd say a note about someone. >> No CRMs on the pod. Banned.
19:27
That's [laughter] how I'm going All right. All right. Let's Let's not say CRM.
19:34
Let's say um the network of people you care about in your life. Okay, fair.
19:37
Uh other people would use it while they were driving to prep for In one case, a professor would use it to prep for a class that she was going to be talking through.
19:45
So a lot of the magic of Stream is conversation.
19:49
It doesn't just capture what you say, it remembers and will actually talk back to you in your own voice.
19:56
So, you can use it as a thinking partner.
19:58
And that's where we saw use cases like you know, planning a lesson plan, prepping for a meeting, talking about an essay you were writing, begin to really blossom off.
20:06
And recently we've been experimenting even farther, which is, you know, now LLMs don't just respond, they can actually act.
20:15
And so, what would it mean if wherever you could you you were, you could fire off an action and and have something start going.
20:21
So, that's a lot of what we've been now playing around with.
20:25
Oh, I didn't even think about that since you started building this, especially in the last like 6 months.
20:30
Yeah, the agent stuff has really taken off.
20:32
Has that just changed the the road map for the app?
20:34
I would say it's it's opened it up quite a lot.
20:39
Um But also like so many temptations, right?
20:42
Every founder just wants to do trip planning for their gadget or their AI, right? Now you technically can. That's true.
20:48
I mean, I'd say we we really try to just start with prototypes we try on ourselves and people close to us and see if it actually is useful.
20:56
So, and we're we're starting with some hypotheses that we hope are generally useful regardless of what form agents or other technologies take.
21:05
When we were starting, we believed that there was no interface to make use of LLMs in the moment.
21:13
And eventually felt it would require some mix of a voice, of personal privacy, of immediacy, um of control over the cadence of the conversation, as well as a lot of work on uh model-centric things like uh latency, memory, personality, tool calling and routing.
21:33
And that's kind of agnostic to what you're trying to do with the AI.
21:36
Uh you need a personality that can present information concisely and helpfully.
21:43
Uh you need memory that can maintain context over multiple conversations.
21:48
And now uh with agents, we see that growing into a new set of primitives that can be layered on that can allow you to talk through and take actions on basically whatever thing you want to use an agent for.
22:00
I saw you tweeted new hardware plus software will finally unlock telepathy.
22:05
Maybe this was a while ago, but it sounds like that's that's the vision here. That's the dream.
22:09
Uh most of my background was in neural interfaces.
22:10
So, it was a bit of a winding path to consumer hardware.
22:15
And for those who don't know, explain neural interfaces cuz it it is kind of as freaky as it sounds.
22:21
For me, neural interfaces were always about expressing yourself fully.
22:25
There are all kinds of thoughts that race through our minds, all kinds of emotions and experiences.
22:31
We have tools to share our full experience with other people like creating art, uh things that can't easily be put into language.
22:40
And I thought neural interfaces could be beautiful because they could allow us to express even more completely than that.
22:49
So, in undergrad, I spent a lot of time doing research at the MIT Media Lab, um and actually briefly interned for Brian Johnson at Kernel when they were doing brain implants, and then moved to New York to join a startup called Control Labs that was developing a peripheral neural interface um for computer control.
23:04
Through that, joined Meta and was working on a lot of uh deep learning models and wearables uh there uh that eventually led to uh Sandbar.
23:13
Yeah, and for people who don't know about Control Labs, uh it was the foundation of the band, the neural band that is connected to Meta's new display glasses, which I have and honestly have been very impressed by it every phase of seeing that band develop over the years um inside Meta.
23:32
Uh and the haptics of it are what's also really incredible.
23:34
Like it it it's not really it's not technically reading your mind, but it's interpreting uh impulses, pulses, right? How would you say it?
23:44
You you can say it better than me.
23:45
It's interpreting pulses from your your muscles in your arm, is that right? That's right.
23:49
So, when you move, your brain sends a signal down your nervous system.
23:55
And when that signal hits a muscle, the muscle amplifies the signal significantly.
24:02
And that creates an electromagnetic field.
24:04
And then we had sensors in a wristband that could read the electromagnetic field.
24:07
And then we used deep learning models trained on really incredible scale to interpret uh those electromagnetic fields into inputs.
24:15
So, you could do something as subtle as writing with your fingertip, you know, on a table, and we could reconstruct um what you had written as an easier way to write on the go.
24:25
Ellis tried this in Vegas.
24:27
Last time we saw you was during CES.
24:29
We met up on the casino floor.
24:29
I have a photo of Ellis wearing the glasses and and using the band.
24:35
Ellis, do you remember it?
24:37
I do, but I mean, I'm just kind of getting getting getting a little spooky feelings right now.
24:40
I'm reading the original Ghost in the Shell manga.
24:45
Have you ever read it, Mina?
24:45
Uh haven't read it, but huge fan of the the movie.
24:49
Yeah, the movie's incredible, and you know, there's just this whole idea that in the future everybody kind of has a cyborg cyberbrain, and a lot of people even have interchangeable bodies.
24:58
And it seems like is that not like what you were literally building toward?
25:02
Perhaps not in terms of ethos, but like if we could read all those electrical signals, you could pass those along to electronic prosthetics, theoretically. How many years out?
25:13
I got to experience some of this, you know, really personally in prior work.
25:16
Uh the lab that I was working in at the Media Lab was focused on neural prostheses.
25:21
So, prosthetic limbs for individuals who had uh perhaps lost a limb and um needed to regain mobility.
25:29
Um before that I had briefly interned at the Applied Physics Lab that was similarly doing neural prosthetics, but for arms.
25:35
I really think of this technology in terms of of giving people freedom.
25:43
Um I think that there is like there are a lot of challenging questions around like what do we want our relationship with technology to be.
25:48
A core belief that I've developed through these experiences is that the best relationship comes when technology acts as if it is an extension of you, rather than a companion or an assistant or a replacement or even necessarily a tool.
26:07
There is a little bit of uncanniness at the start of that, which is, well, what should my relationship to this self-like thing uh be? Is it me? Is it not be?
26:15
And I think that's where some of the perhaps what you're describing, Ellis, um that question comes from.
26:24
But when something has your context, acts with your goals, and is a literal extension of what you want to have happen, I think you eliminate a lot of the gap that comes with something that is other than you that creates a tension where perhaps the goals are misaligned or perhaps the the context isn't quite right and you're having to overexplain yourself.
26:45
Um And so, to me, whether it be uh a ring or a neural interface, the goal really is about letting you express yourself very easily.
26:53
That's come through very clearly in the product, I think.
26:55
And like when you showed it to me, uh God, whenever that was, that was a while ago.
27:01
And Ellis feel us and I love you brought the case with all the different rings.
27:05
And then we were talking about it in Vegas with Ellis.
27:06
I mean, the way you've designed it, the like intentionality of you have to hold it up almost like you're, you know, a secret service member like talking into their sleeve, right?
27:15
And prompt it to engage with you.
27:18
And then you have that haptic feedback of it vibrates to let you know, right?
27:21
That it's listening versus being like always on or chiming in, which is even more assertive like for an AI.
27:27
And a lot of hardware products are going that direction.
27:30
Um it seems like you've clearly viewing you're viewing the ring as like something that the human prompts and fully controls.
27:39
Yes, and I think that having a strong belief um or a north star really helps with product design, whether software or hardware or interfaces.
27:50
And I'll often chat with other founders building software or hardware who are perhaps struggling to make decisions one way or another.
27:59
And I think if you can plant a flag and say, "This is what I believe in."
28:02
and allow that flag to dictate choices to you, it makes a lot of things much easier.
28:10
You know, for example, putting haptics into the ring um was not easy.
28:16
>> [laughter] >> Um but we knew in order to give that feeling of something being an extension of you, uh we would need to tap into many more of the body's senses.
28:26
And providing haptics is a great way of closing the loop that what you have said has been heard, and potentially even silently prompting you if, for example, a reminder is due, or perhaps in the future if, you know, some action taking place needs a clarification.
28:41
And similarly, uh originally, Streams' voice was just an off-the-shelf voice.
28:47
Uh but it didn't feel quite right.
28:49
And I remember Croc said to me, "If you're serious about this to being an extension of you, it should be your own voice."
28:58
Uh and I thought, "Oh, it's going to be so weird." Uh and we tried it out.
28:59
And for the first day, it was kind of uncanny.
29:03
And then by the second day, uh everyone on the on the team just really loved that being their inner voice.
29:10
And as we've continued to roll it out, we've been surprised at how many people seem to uh prefer it.
29:15
I need to use it more, but that kind of freaks me out. I'm not going to lie.
29:18
When you had it talk like me.
29:21
Do you guys hear your voice in your head when you think or plan things?
29:24
So, I don't think I hear my voice. Do some people? Is that a thing?
29:28
So, I know some people can't visualize things in their head.
29:31
I think it's like maybe 30% or so.
29:35
I might be getting that stat wrong, but um don't hear, and then a similar percent perhaps don't see, and then some people don't see or hear in their head.
29:43
Um Ultimately, self-extension dictates a lot of things to us about our direction.
29:48
One is that that default voice is an inner voice, but another core value of self-extension is is self-autonomy.
29:53
You should get to make choices and we as a company should dictate as few of those choices to you as possible.
30:01
So, you'll always be able to swap off to another voice uh if you, you know, prefer that.
30:09
As long as it's not my grandma's Waze voice that she uses, which is like the voice of a British butler who's like, "Please make a U-turn promptly."
30:20
And it's kind of gravelly.
30:20
Uh just do better than that.
30:22
That's that's not that's not the pick for you. Not the pick for me.
30:25
That doesn't help me do my best thinking.
30:28
Um but uh I mean, speaking of kind of like I mean, there's clearly such a naturalistic vibe to what you're building, and one of the things I often struggle with personally is when I'm walking outside or sitting outside, you know, we all have the tendency to like pick up our phone and see what's going on or check our notifications or whatever.
30:46
Um but I've been trying to like actively not use my phone while walking and just kind of taking the sights.
30:51
How do you guys think about that as you make technology even more readily accessible?
30:58
And I mean, obviously, you know, it's not any more accessible necessarily than something like an Apple Watch, but it's clear that you want to maybe accelerate this idea that people are putting their ring up to their mouth and logging some thoughts while they're walking or being outside.
31:11
I honestly love how in flow I get to be using stream.
31:13
A lot of the initial motivation to even work on this problem was that I walk everywhere in New York, and I wanted a way to capture thoughts or talk through ideas uh without interrupting.
31:26
And for me, a phone wasn't quite right because you'd have to stop and pull it out.
31:32
Um earbuds weren't quite quite right because there's really no easy way to invoke them, and it's kind of embarrassing to project out into the void.
31:42
No, you say, "Hey Siri, please take a note."
31:45
You want to call your ex? Dialing now.
31:50
>> [laughter] >> You know, actually, it's uh that reminds me of um we didn't intend entirely like top-down intend this from the start, but uh a strong intuition we had early on was that um the ring would need to be capable of both conversation and music control.
32:08
Because if you're to do that interaction you just described, what you'd immediately then want to do is resume whatever music you were listening to.
32:15
So, the nice thing about stream is, you know, I could, for example, say, um "Can you tell me a little bit about um the Access podcast?"
32:24
>> Sorry, I'm having a connection issue.
32:25
Intruder in the question.
32:26
>> [laughter] >> Intruder in the question.
32:27
You try that one more time.
32:30
That was perfect comedic timing.
32:33
>> [laughter] >> You know it's the real demo when it bugs.
32:35
I wish it was in your voice though as well. Uh it should be.
32:37
That's uh >> [laughter] >> Can you hear me?
32:42
>> [music] >> There we go. Is the volume clear? You doing okay?
32:46
Can you tell me about the uh Access podcast?
32:51
Yeah, Access is a tech industry podcast hosted by Alex Heath and Ellis Hamburger.
32:54
They interview big players and emerging founders, kind of the inside baseball of Silicon Valley.
32:58
And then if I wanted to resume whatever I was listening to, I could just tap again, and then it'll resume.
33:05
That might not come through on the mic, but I can also swipe to control my volume, and that lets me just stay in flow if I had been walking or biking or driving uh and listening to a podcast or a song like Access.
33:18
Okay, this product is dope.
33:21
This is too good to pass up.
33:21
Can you ask it what you've been worried about lately?
33:23
Uh so, this is actually a demo account because I intentionally am too embarrassed to show my actual stream.
33:31
Um it will pull up things from a synthetic Mina's memories.
33:33
So, if you zoom out, you can see things in the past.
33:38
Um we can see what it it thinks the synthetic Mina was worried about.
33:44
Can you tell me some of what I've been thinking about uh and maybe worried about recently?
33:49
Yeah, from how our talks have gone, a few threads run through things.
33:51
You've been circling around creative authenticity, pressure to perform, needing quiet trust in your own pace, and wanting steadier ground beneath the work.
34:00
Does that fit Okay, so this is the ultimate performative male.
34:04
If you've heard of the performative male means on the subway, uh a thinkbro narrating his personal inner monologue into his ring.
34:10
Ask it to prep you for this cuz you asked it about the podcast.
34:14
Say, "Can you prep me to be on it?"
34:16
I'm about to go on the Access podcast.
34:19
Some might say I'm already on it.
34:20
Uh but can you help me figure out how to prep and maybe even talk about you? That's a cool moment.
34:27
If you mean literally tap, just a single tap on the ring will wake or pause me.
34:30
For talking about me, what angle are you thinking?
34:34
Tech design, philosophy, or relationship? I see.
34:36
So, it's it's pretty steerable.
34:38
It's not like going to come at you with like a huge long thing.
34:42
It's going to like ask you for how to guide it. That's right.
34:46
It's uh that's a lot of the work we do on its personality.
34:50
We want it to present information very clearly, and we want to have it ask questions that get you thinking.
34:56
Um but otherwise, uh what we say is we want personality without identity.
35:01
It shouldn't have opinions. It shouldn't have goals.
35:05
Ultimately, down the line, it should be alterable by you if you would like to give it those things, but it turns out it's actually relatively hard to get a model to behave that way out the box.
35:12
Uh so, we've had to build a lot of infrastructure for creating like in this account, synthetic users that stream interacts with, and then we have judges that basically look at interactions with respect to the goals we have for the personality.
35:25
Since the app has gotten more full-featured, are you using ChatGPT or and/or Claude uh less? That's a good question.
35:37
I've noticed on the go, effectively all of my usage has gone onto stream, and I no longer use other apps uh like AI apps or notes apps.
35:50
Uh on desktop, I can I continue to use them definitely.
35:53
Um and I think ultimately, our goal really isn't to compete with ChatGPT or Claude.
35:57
We're trying to present a relatively flexible interface at the end of the day, and we're starting with some specific things that you can't really easily do on the go that hopefully we can offer to you more easily.
36:11
Um but if you'd like to use this interface in other ways, as a direct input to those AIs or or for other applications like that also comes from our self-extension kind of value dictating back to us.
36:26
Um so, for example, we recently shared um uh a sign-up form for our private SDK alpha, where we're looking for developers that would like to try building on top of stream.
36:36
So, uh we can basically develop that SDK together and let other people use this how they will.
36:43
Could you do an MCP type thing and just let it pipe into my Claude that way or whatever?
36:48
Could I just connect via MCP?
36:51
I think the short answer is uh yes, but a lot of things break really quickly.
36:56
There are issues with context management, conversation management, latency, asynchrony, a few other things that MCP doesn't solve out of the box uh that we're trying to make uh ideally uh easier for people to to plug and play with.
37:13
I'm picturing in a couple years' time Alex speed walking through the airport talking to his stream ring.
37:20
"Hey, uh mirror world Alex, water the plants, walk the dogs, make dinner for my life my lovely wife. Make no mistakes."
37:26
I mean, I am going to use the out of this thing.
37:31
Like, I cannot tell you how often I am like, "Damn, I wish I had this ring on."
37:36
Ever since you first showed it to me, Mina, it's been one of those things that consistently comes up for me in a way that most demos don't.
37:44
Um I'm just like, "This is such a natural way to quickly like remember something, jog a note, prompt, do a quick prompt, right?"
37:51
Like, until we have glasses or whatever or, you know, I don't know if the necklace wearable thing will ever really take off, but until we have some kind of easier input into AI that we can wear, um there's just all this friction.
38:04
And it's not even about the AI piece, honestly.
38:08
It's like memory recall for me.
38:10
It's like there's so many times where I'm context switching or I'm walking out of a meeting or an interview, and I want to or walking the dogs and I want to like remember something, and you just don't want to type it, you know, you know?
38:21
Like, even with WhisperFlow lately, I've just been dictating more, and that's been a help, but man, this is like >> Wait, I thought you were using monologue.
38:27
Are you already on to the next thing?
38:30
I switched to WhisperFlow.
38:30
You have to tell me when these things happen.
38:32
All respect to Dan and the Every team, but uh yeah, I I think WhisperFlow is is is my my jam now.
38:38
We in many ways I think lucked out with the timing of voice as an interface. Uh So funny.
38:46
I feel like people have been talking about it for 15 years. Finally, it's here.
38:53
>> It's, you know, all all things are converging.
38:55
All all roads are roaming.
38:58
Um One one thing that I would say that I haven't seen as much of though is a distinction between voice and conversation.
39:08
And I would say we we believe in voice, but we believe even more in conversation.
39:12
Because something we've observed is a lot of people see what we're building and say, "Oh, I would love to use this to capture.
39:18
I would love to use this as a voice input device."
39:23
But living with it, what you rapidly find is sometimes you need to recall something.
39:30
Sometimes you need a little bit more information, uh especially if you're on the go.
39:34
Uh sometimes that thing that you're capturing, it's not just a one-and-done thing.
39:38
You're actually iterating on an idea.
39:41
And that's where having something be able to speak back to you becomes so important.
39:45
You can I think see that in most UIs today.
39:47
There are very few one-way interfaces.
39:50
You know, Cloud Code will ask for clarifications as it's building.
39:55
Um and so you'll need some way of delivering output back to the user.
40:01
And that's where a lot of the hardest feedback and ML and infrastructure problems uh end up being.
40:06
I mean, it's got to be so tempting kind of like what I was saying to just go the mega agnostic utility route, but it's clear that your heart is in a very specific place with the type of life you want to afford people.
40:19
And I would also imagine that imminently, I mean, because you've been able to build it, you know, the factories are out there and there's going to be a flood of like AI voice capture clones from China hitting Temu and Amazon and everything within a year or two in all likelihood.
40:36
I mean, does that mean does that make it easier to kind of choose a narrower path or do you want to win that battle, too, you think?
40:45
I would I I would say we have a lot of ambition.
40:48
>> [laughter] >> Um uh I think at the end of the day we are trying to create a new interaction model that becomes the primary interaction model.
40:56
You know, we'll see whether we get there, but I think getting there depends on making things that are useful, that people like.
41:05
And so I definitely watch other things that are happening.
41:09
Um but none of us have proven that this is the right thing.
41:13
And I think that's still the obligation our team has is to like not get distracted, just make sure it's like as useful as beneficial as it can be.
41:24
You know, I have a lot of faith in our team's ability to to to do that at high quality um both on the hardware and the software side.
41:30
Um I mean, hardware alone, there is a lot that goes into making something that is very reliable and very safe.
41:36
Uh you know uh we have a pretty stacked team.
41:40
Our head of hardware was the head of hardware for Fitbit and Peloton and made the Kindle.
41:46
Uh we have uh many engineers from you know, Apple and from uh uh uh from elsewhere that uh are oftentimes focused, for example, purely on the battery or focused purely on um sensor fusion for the inputs to the to the ring.
42:04
And on the software side, we're similarly working to build effectively everything down to not down to right above the foundation model from scratch.
42:15
Uh so all of the infra and connectivity and iOS experience and evals and uh tuning.
42:20
Some of those things I think will be easier to replicate than others.
42:25
At the end of the day all that matters is is it a good product.
42:29
Is that haptics thing patent pending?
42:32
I would say we are uh we we have a lot of IP in motion.
42:37
Does anybody even do that anymore, Alex?
42:40
Patents for hardware matter.
42:40
I don't think they really matter for software, right?
42:43
>> Enforcing patents, I mean, you want to spend 10 years in court against Samsung or whatever and spend all your VC money on that? Hell yeah, you do.
42:50
We like to de de devote most of our time to building good products.
42:55
Um but uh we've also been working on this for for for a long time and have learned a lot in the process.
43:01
I know you've got this launch uh of the ring, obviously, that you've been working towards, but my understanding is you're thinking about other form factors.
43:08
Is there other types of wearables that you think are going to actually work?
43:13
I mean, everyone's trying rings, bracelets, necklaces, pendants, pucks, right?
43:20
Open AI is probably going to do some kind of puck type thing.
43:23
Um what do you think is the other compelling form factor out there? Glasses?
43:27
Yeah, I think I think about it almost in terms of like what parts of the IO challenge are you uh solving or are you missing?
43:37
Um when we started Sandbar, we were building a few things in parallel and trying them out.
43:42
And what we arrived at was a feeling of we think the ring solves voice input.
43:47
It's very fast, it's very private.
43:50
Uh it allows you to do a push-to-talk control um interaction.
43:52
Um but it doesn't solve other things.
43:55
It doesn't serve cer- uh solve certain audio contexts.
43:59
Um it doesn't solve vision.
44:01
Frankly, if I like had a confident belief in what that thing was, I would say like, oh yeah, and we're working on XYZ.
44:08
And we we just haven't gotten there.
44:11
Um we've tried a lot of things, but those things haven't crossed our bar.
44:15
Um all I can say is like, I do believe that there will be uh other form factors necessary to solve other parts of the interaction.
44:23
It's like the main thing I miss of Apple under Steve Jobs is that he had a bar, right?
44:29
And they test so many things internally as you are doing and you clearly have a bar for when you feel like it's ready.
44:35
And if you do believe that you understand consumer behavior and you have better taste than anybody else, you ship it exactly when you know it's ready, you know, like how on the iPhone was later than the Treo and the Pocket PC and all that.
44:47
And to me, that's kind of the most disappointing aspect of like the Vision Pro in hindsight is that it showed that they uh decided to kind of go with what Wall Street wanted from them, even knowing full well, I would think internally, that most consumers are not ready for this quite yet.
45:06
It's not light enough quite yet.
45:08
The software isn't there quite yet.
45:09
And as soon as it's ready, they can come out with the world's best hardware for it.
45:12
And um yeah, that's my soapbox of the day.
45:16
I go back and forth on that.
45:18
Um I think you learn a lot from things out in the world.
45:22
And so I I'd probably take the other side that I think that they may have released at the right time.
45:27
Um like they they they demonstrated a lot of new-to-world capabilities.
45:31
Um they got to see a lot of use cases.
45:34
Um they got to share their interpretation of what that future could look like.
45:40
And it typically takes a few revisions for um any product, but especially for a consumer hardware product, to become mass market.
45:50
Um you know, I I forgot >> Isn't it the third one?
45:53
It's always like the third one is the one that's supposed to hit. Yeah.
45:56
Yeah, I I I I don't remember which Quest model really got mass market or or had a mass >> the third one and it was like same thing with the iPhone.
46:05
It wasn't until like 3G that it like really hockey sticked like That kind of that's a that's a pattern with consumer hardware, I think. >> I know that story. I don't I don't know.
46:17
People freaking loved their OG iPhones and iPods and AirPods.
46:22
No, I'm just talking about like the companies that make the hardware feeling good about the product.
46:26
It usually takes getting to gen 3 is the thing you hear time and time again cuz you learn all the errors in one, you refine them in two, and then three is when you really get to like build on that and and ship the thing that you're proud of. Is that fair, Mina?
46:40
I I I think that's that's true, but I guess maybe what I'm hearing from you, Ellis, is um a distinction between a product becoming mature versus a a product being right. Right.
46:50
I mean, it seems that they were caught off guard by the response and a lot of the commercials were fake.
46:59
You saw this with Apple Intelligence as well with the fake commercials that they pulled.
47:03
And they just like weren't realistic in any way.
47:07
And I mean, if you look at the OG iPhone, they thought it was good enough to be out there.
47:10
They knew it didn't have 3G.
47:11
They knew it didn't have copy and paste.
47:13
That was very clearly on the road map.
47:15
And it felt like they had their ideas in order about a very strong conviction about, you know, what the future was going to look like.
47:21
And I can see you guys in your office with the Apple people, the Fitbit people, being like, well, these glasses are cool, but they're still too big.
47:29
And so that's that until it's smaller, you know.
47:31
And that's why I think the Ray-Ban Meta thing has been so successful.
47:35
I mean, we had Alex from from Meta, who you know, on the show and he was like the people at Ray-Ban were telling us, giving us super hard constraints that they wouldn't have otherwise had.
47:46
And I do think that Apple used to have that type of bar for like what might be considered, you know, culturally acceptable.
47:54
Um that maybe isn't quite where it used to be. I don't know.
47:57
I think one benefit that we had from the start was like how unimportant we are. Uh And totally.
48:05
also being in New York, like we were wearing this and way crazier prototypes just in public from the very start.
48:12
And no one minded it partly cuz it's New York, partly because we're unimportant.
48:16
And a lot of our bar is our own feeling using the the prototype or the product.
48:22
But a lot of our bar is watching and seeing how other people react.
48:26
Um and and frankly what got us over the bar to saying we should ship the ring um was how many people were telling us uh like, sell me this. Take my money.
48:41
Always a good sign for a startup.
48:41
And and so we were thinking, okay, that that probably means it's it's directionally um uh uh correct.
48:50
So where do you think wearables have gone right and gone wrong uh over the last several years?
48:56
I think one thing that comes to mind, even though Apple has attempted to tamp it down, is just how much of a distraction notifications on Apple Watch are.
49:05
Like and I think a lot of people have started to push back on that and even stopped wearing them as a result.
49:11
They're also just incredibly lame, Ellis.
49:13
I mean, I think that's why >> [laughter] >> But yeah. Yeah.
49:18
What do you think about, Mina, in that department?
49:21
One thing I find really interesting is that we've effectively never seen uh input or interface wearables.
49:28
Uh you could maybe debate Apple Watch, but it's primarily a health product.
49:34
Um as are effectively all other successful wearables.
49:40
Um, of course, there's hearables and headphones that are really focused on music.
49:45
But, in terms of wearables that give you new computing experiences, we we really haven't seen anything like that.
49:55
Um, I think the meta display and the neural band was perhaps the first step in that direction.
50:03
And this is something we talk a lot about internally.
50:05
We live with stream every day.
50:08
So, we get to feel how intuitive and familiar it is.
50:10
But, the even fact that you can talk into a ring, uh, is like so outside the realm of what most people have ever seen, um, that's going to take a while, uh, for people to get used to that new reality.
50:23
So, I do think that any new wearable needs to spend a lot of time on storytelling, on user education, on making things really accessible.
50:35
Otherwise, they just remain kind of tech gadgets or or or toys for like a very select few who are willing to jump through the hoops.
50:45
>> We need a air horn sound for whenever a guest says storytelling.
50:49
>> [laughter] >> Cuz then that activates Ellis.
50:52
Yeah, it it is interesting though.
50:52
It is such a positioning thing, isn't it?
50:54
I mean, I think it's easy to forget that the Apple Watch was not initially positioned as a health product.
51:01
And I feel like, I mean, that actually may play into your narrative.
51:05
Uh, that it may have been gen 3 or around then that they started repositioning to health and that clearly was was, uh, the right approach.
51:13
But, I mean, I know they also, I don't know if anybody uses it, but doesn't Siri activate when you raise to wake?
51:19
Uh, I don't know if it's by default or not, but There's no ability to control when Siri activates, but I think you're >> [laughter] >> It's spontaneous.
51:31
That that does make me think about how when you look at a wearable or you look at a hardware product, um, it's easy to miss how much software is required for that experience to make any sense or feel good.
51:47
The parallel I would give is uh, if I handed you an iPhone 1, um, and it had none of the software required to drive multi-touch or do app switching and pinch to zoom and the background data management, um, what good would it be?
52:04
Uh, if I were to give you a mouse and a keyboard and a monitor, uh, but I couldn't do, um, windowing, um, I couldn't do clickable icons, uh, I couldn't do all of the other things that a graphical user interface affords, it would similarly not be a good product.
52:21
Um, and so, we've really tried to treat interface design as, um, everything from the thought popping your head to whatever you want next.
52:31
Uh, and that's where things like conversation, personality, memory and retrieval, um, very low latency, asynchronous parallel processing, etc. , etc.
52:42
, buzzwords, uh, become so important.
52:44
Uh, and ends up being where we spend a huge amount of our time.
52:47
Do you think about, you know, I just saw Oura, for example, they're going to IPO probably this year.
52:53
Do you think about someone like that who's approached it from the health angle, uh, the form factor, the ring form factor, potentially trying to do what you're doing?
53:02
Is that competition, uh, something like an Oura?
53:04
I think the honest answer is I don't really know.
53:09
Um, I'm curious to see, but at the end of the day, like, I really deeply believe that what you believe in dictates the choices you make.
53:17
Ideally, you know what you believe in, um, otherwise, you'll be making choices without understanding the the deep why.
53:23
I think for most wearable companies, their why is really rooted in health.
53:29
So, even if they're adding new features, I expect that to really be the focus.
53:35
Uh, we are really focused on human-computer interaction.
53:37
We are trying to help you think, remember, act, uh, wherever you are.
53:43
Um, and so, uh, unless someone has the same mission as us, I doubt they'll arrive at the same conclusion.
53:52
I like that because all these health wearables, uh, they they suck.
53:56
I just stopped using them.
53:56
Like, I I've I'm not going to use the Oura anymore.
54:00
I got a Whoop like a month ago and I'm already not wearing it.
54:04
I mean, Ellis, I feel like you've talked about this, too.
54:05
Like, these health wearables kind of exist to just make you feel bad about yourself.
54:09
Like, oh, like, you you woke up tired and guess what?
54:14
Your sleep score is a 43. It makes you feel worse. Yeah.
54:18
>> [laughter] >> I mean, I I'm curious, I mean, I mean, it's clear, you know, whether it's the the telepathy or kind of the ideation, you are trying to afford people this ability to have bigger thoughts and think clearer.
54:29
Do you have a sense of kind of what those thoughts and ideas will be?
54:33
I feel like there's a pretty big hot button issue right now about like what ideas are even going to matter that people have in the coming years, you know what I mean?
54:42
Uh, there's also a world where you just make it about recording your groceries as you remember them at the fridge, which I know is also a feature that you have.
54:49
But, um, are you kind of hoping people are going to have these these bigger ideas that you're helping string together with the software as you said?
54:57
How woo do you want to get? Oh. Well, go all the way. Follow your heart. Follow your ring finger.
55:04
Technically, it's index finger, but, um, >> [laughter] >> um, one thing I really believe is that all thoughts are beautiful.
55:10
Whether it's a grocery list, whether it's a heartfelt letter, whether it's some important interview you're prepping for, um, whether it's a cloud code instance you're trying to check on in your Mac mini, the fact that you are having those thoughts, I think, is what makes it beautiful.
55:29
And what we are trying to do is make an interface that doesn't operate under the false assumption that there is a division in our minds between the practical and the personal and the professional.
55:40
Um, the nice thing in my mind about existing note and conversation tools is that they are the most flexible interfaces we have.
55:48
And so, you can use them in all of these ways.
55:52
But, I think this becomes so important as we step into the future.
55:58
I remember I started working, uh, with LLMs in like '21, '22.
56:01
And then, ChatGPT came out and there was a lot of excitement, but there was also a lot of concern about what the future would look like.
56:11
And something that was very important to me was I wanted to build something that would be valuable when we reached general intelligence.
56:19
And to me, that looked like meeting people where they are.
56:21
Even after general intelligence, we'll be taking walks, we'll be thinking, we'll be remembering, we'll have things that we need to do.
56:30
And importantly, I wanted to build something that prioritized and extended and augmented humans.
56:38
Because I think we we we ought to experience the freedom that this new technology affords.
56:42
One could say a bicycle for the mind. Oh, god. God, don't. Also banned. No CRMs on the show.
56:50
No bicycle for the mind on the show.
56:52
Uh, I I guess what I will finish that with is is what I feel pretty palpably day-to-day in Trace what it's like in in in LA, but definitely in New York, I feel people's fear of replacement and fear of how this technology is being portrayed.
57:09
And I don't really see the human or extension at the core of many of these narratives.
57:15
And I do believe it's a choice, and that choice comes from the relationship that you design into the product and into the technology.
57:21
And we can choose to make these things an extension of us.
57:24
You know, if we do anything, I I hope that that is a message that we can contribute to pushing forward at least a bit.
57:32
So, you're hoping people don't deface your billboards.
57:33
I hope people can take more walks and feel good about themselves.
57:39
>> [laughter] >> Uh, uh, uh, I I hope people can look forward to the future.
57:43
I hope people can wake up and be really happy.
57:45
I'm certainly not worried about like the art and expression worlds because I think that this is a tool that a lot of people are going to be able to make a lot more with, uh, versus people making less with relatively.
58:01
Um, and I think there's a very big difference be- behind I think there's a very big difference, uh, between a song made with Suno that is intended to just kind of play as background music and doesn't really have an opinion or your lived experiences and thoughts within it.
58:21
There's a big difference between that and a song on Suno that you might see that has like an artist statement about what the person was going for with it.
58:28
Um, you know, like, my friend was able to make a song about his dog, uh, that he never would have been able to make before. Like, great.
58:36
And it's hilarious because he made it and it was a side of his expression that I'd never seen before.
58:42
Um, but I do think, you know, there there is going to have to be some retraining or or different expectations for peoples of themselves to like kind of free them to think in this way, you know, outside of your current lane that that the world wants to put you in.
58:57
I mean, I have I've definitely seen the beginnings of that.
59:01
I've seen people who are creating at a volume that, you know, it is is stunning, um, as well as people who are in creative ruts for a long time.
59:11
At the end of the day, I personally don't have a lot of, um, judgments or opinions on how people like should conduct their lives.
59:20
Like, I'm not expecting everyone go off on like Zen, um, meditations and be thinking deep thoughts 24/7 and etc. , etc.
59:27
I do think people would benefit from a life that generally affords them more flexibility and freedom to move and think as they prefer or as they choose or rather than as they'd have to.
59:40
Alex, when you've automated everything, is what's left just like your relationships? Yeah.
59:48
I mean, yeah, it's interesting.
59:48
Me and I've I've talked a little bit about this, but I I don't really write anymore, like my stories. I am I talk to AI.
1:00:00
And it has all my contacts and I talk to it like an editor talking to a reporter crafting a story.
1:00:05
So, I say, "No, that's not good.
1:00:07
That's a lazy transition.
1:00:08
It's smarter to lead with this angle cuz this is fresh. Move this graph here. Your kicker sucks."
1:00:13
You know, like like I'm like I'm training a young reporter, except it's hopefully my AI that gets smarter.
1:00:20
Although, I will say uh Claude and open AI, need to work on that memory thing. It's not very good.
1:00:28
So, it's really changed the zero-to-one creation process for me and and made it more fun to actually like put writing out cuz I always hated that.
1:00:36
I always hated just like looking at a blank page and a bunch of transcripts of interviews and needing to assemble all that and then begin, you know, the process.
1:00:44
And this gets me there like immediately.
1:00:46
And it does feel empowering.
1:00:48
So, are we moving to a future where you have schmooze leverage and that's all you have? Schmooze leverage?
1:00:53
Yeah, and like taste leverage to go like >> [laughter] >> ideally, ideally, this is not happening right now, but ideally what this does is it frees me up because my writing process takes less time to consume more, to talk to more people, to build more of a bird's-eye view of what I'm covering.
1:01:10
And then, what you're really getting is like my curation of this universe that I'm covering and access to the people I'm talking to through the pod, through the newsletter.
1:01:20
And that to me is my edge, not my prose.
1:01:25
I think I I've been talking about this a lot, but I think writing will generally go towards very human anti-AI beautiful prose that honestly gets weirder and weirder and more creative, pushing the boundaries.
1:01:36
And then, writing that is just kind of a um a medium or a means to something else, which is like in my case reporting and some other case maybe it's like news summary or whatever.
1:01:50
And I think everything else in the middle is going to have a really hard time.
1:01:53
That's like my working theory right now. I would buy that.
1:01:55
I mean, you're each pretty great people to hang out with and that interpersonal I think dynamic really matters.
1:02:01
I think um that's part of why I especially like when this technology is framed as self rather than other because when AI is other and is competing with other humans, it almost confuses what the actual benefit is of someone who is different than you.
1:02:18
Like a lot of why I enjoy talking with you or I enjoy talking with other people is other people challenge me and they have lived experiences that are very different than my own.
1:02:30
And those things are pretty hard to get from a machine, but you can kind of be fooled every now and then, which which I don't necessarily think is is helpful to people.
1:02:40
The other thing you're saying that makes me think of is something that Kirat told me when we started working together on this was how he wanted to program while walking through a forest and he wondered if what we were building could enable that.
1:02:56
And similarly, I wonder what it would look like if you could just take a walk and write a a a story.
1:03:02
And whether that would change what the story would look like because you'd be thinking with your mind and you'd be moving your body.
1:03:09
And um I'll I'll I'll I'll flag that I don't expect again, you know, all of our thinking to occur while in some idyllic motion through the the the wilderness, but I do want people to have more freedom to pick that and to use the interfaces they want wherever they are in whatever ergonomic or bodily positions that they would prefer to be in rather than having to be chained to whatever you know, window we find ourselves in.
1:03:38
So, tell us how has AI changed how you built and organized the company?
1:03:44
I have one funny thing I was thinking of, which is how design crits go when someone's idea is no good, everybody says it's no good and then they say, "But all thoughts are beautiful."
1:03:55
>> [laughter] >> How does the company run with AI differently?
1:03:59
I can say perhaps how I run and then how I see our our team running.
1:04:04
I think the two biggest impacts to me is I arrive at decisions much more quickly and I'm creating a lot more.
1:04:15
So, I use usually now either stream or Claude, but sometimes chat um to think through things that might otherwise have required me to like write and ponder and you know, internalize for for for very long.
1:04:31
These things are just good at asking me questions that route me to my true intention pretty pretty quickly.
1:04:37
The other thing is that like many, you know, founder CEO types, most of my life is like some mix of hiring, fundraising, etc.
1:04:48
But now that I can uh you know, build with Claude code and I can create images with image models um there was a good period where like many other people I was just like sleeping very little and just making a lot and then eventually had to be like, "No, you need to be resting more, Mina."
1:05:03
Um but for example, there are some images that we've been posting recently on Instagram that I I I absolutely love.
1:05:09
They're like black and white images of the ring balancing on strange body parts.
1:05:12
Um and those I was able to make um using AI image tools.
1:05:19
And I'm seeing something pretty similarly with the team.
1:05:21
I think a lot of people are talking about the impact of AI on coding.
1:05:28
But I'll say it's impacting everything for us.
1:05:29
Uh we do everything in-house including marketing, supply chain, logistics, ops, etc.
1:05:36
And we had a a supply chain review a couple of weeks ago where our head of ops basically made this massive reconfigurable dashboard that showed the lead time of every item that lets you switch between you know, this is what different tariffs will result at these different price points.
1:05:58
And we had to make some really complicated decisions that this tool made really easy.
1:06:02
And our head of hardware watching this was stunned.
1:06:06
He was like, "This would have taken a team of people weeks to build at his prior companies at Fitbit and at Peloton."
1:06:16
And now just our head of ops was able to build it himself.
1:06:18
And and he he emphasized that it was very hard to build.
1:06:22
It required a ton of going back and forth with Claude to create that dashboard.
1:06:25
But we all had this collective moment of like um wow, you can you can do so much more across everything.
1:06:32
I have a good question to close on.
1:06:34
Is Bluetooth finally not bad?
1:06:39
Cuz I feel like there were so many gadgets over the years that were trying to do like ambient awareness or, you know, speaking to it or like taking notes, like whatever it is, but Bluetooth was always the constraint because it was so terrible. Is it better now?
1:06:55
It's definitely matured a lot, but it remains pretty hard.
1:06:58
For example, getting to something that is always connected uh is is pretty non-trivial still.
1:07:05
We have two full-time firmware engineers and antenna consultants and connectivity consultants. >> Oh, yeah. You're holding it wrong.
1:07:15
Uh well, >> [laughter] >> it's actually kind of cool seeing the tests that you have to do cuz like you need to check, you know, how how does the connectivity change based on how many people are in a room, how much interference is in a room.
1:07:29
So, the underlying protocol is is is pretty good, but um for better or worse, a lot of hardware in general has not necessarily become that much easier.
1:07:43
I would say in our case we actually have some unique challenges, which is um compared to other health wearables that are just sending your heart rate, we're sending audio.
1:07:52
We're sending a lot more data through a metal housing.
1:07:57
So, the connectivity becomes even more difficult.
1:08:00
Um and that is why that is why we are are taking our time to to ship >> [laughter] >> to make sure that >> what you're most jealous of, their U or W chip or whatever it is? Which one? From Apple. That's a good question.
1:08:15
Um Cuz that seems like one of their key advantages with all this wireless stuff that they've done is just building that whole communication pipeline in-house.
1:08:23
Yeah, I I'd say that's that's probably part of it.
1:08:27
Um Apple does get to do certain things with their devices that aren't immediately available to other peripherals on iOS.
1:08:36
But we're pretty happy with the access that we have, frankly.
1:08:37
Like it's it's enough for our application.
1:08:39
Well, no one thought that was going to be the last question, but my I mean, been in technology a long time and I think Bluetooth and Wi-Fi are my two least favorite technologies. So, we had to end there.
1:08:51
I had to know what the latest status was on those so I could properly monitor the situation. Very important.
1:08:58
Yeah, so thanks for the info.
1:08:58
Yeah, Mina, we really appreciate your time, man.
1:09:01
Looking forward to to getting a ring.
1:09:03
Excited to get you a Yeah, if it isn't completely clear, Alex like really really wants one like as soon as possible.
1:09:11
Like he he's trying to be polite, but I'm really not. >> batch one.
1:09:15
Is that We're going to get you a beta unit, but I think you got in batch one, right? Uh I'm not sure.
1:09:17
We'll figure that out if those are Batch two is uh still shipping within the summer. Um so, uh Okay.
1:09:24
Any listeners who want to get in before uh the next batch. Got it. All right, Mina. We'll let you go. Appreciate it. >> Yeah, thank you.
1:09:35
>> [music] >> And that's it for this week's show.
1:09:38
Don't forget to like, subscribe everywhere you get podcasts. We are access.
1:09:42
show on the internet and you can find us in video at accesspod on YouTube.
1:09:47
And if you like this episode, leave us a five-star review, whatever the max amount of stars are on the platform that you're using, or a thumbs up.
1:09:56
And please share this with a friend, share it on your social networks. Uh really means a lot.
1:10:00
We're a new show and we we love the word of mouth.
1:10:01
As long as five stars have been around, people only ever give one star or five star.
1:10:06
[music] Well, give Access five stars.
1:10:08
I just love that like user interface debate.
1:10:13
It's like, let's just go with the with the thumbs up or thumbs down or something.
1:10:16
Okay, Ellis, we're doing the outro. Come on. >> Okay.
1:10:18
>> [laughter] >> You can find me >> my news You can find Alex's really lovely newsletter at sources. news.
1:10:27
And you can find me at hamburger on Twitter and at meaning. company.
1:10:31
Access is part of the Vox Media Podcast Network and the show is produced by Hooks Creators. Bye. Bye-bye.