Why the next AI boom is physical AI | Caitlin Kalinowski (ex-OpenAI, Meta, Apple)

[00:00] There's a dawning realization, especially in the lab, the acceleration is going so vertical that what you can do behind a keyboard with AI is going to saturate. When that happens, the next frontier is the physical world. Robotics, manufacturing, industrialization. You're living in the future and designing it. There's probably more change in war than there is in consumer electronics in the next two years. We need to invest a lot more in drones than in aircraft carriers. Just imagine 100,000 drones coming out of China just at us. [00:30] that we need to re-industrialize the country significantly to be safe in a military sense.

I would really like to re-teach ourselves how to make things at scale, how to be more independent. People that are your allies now may not be in the future. You worked with some of the most legendary, successful builders. Steve Jobs, Mark Zuckerberg, Sam Altman. Sam is really good at saying, why not more? Why not 100x or 10,000x? You're thinking too small. [01:00] excellence was not wavering. What does it take to create a robot that feels human and connected? If you walk into a room and a robot's just like, like it's creepy.

You want these devices to be non-threatening, appear soft, reactive to you. Pixar, Disney are probably the world's best at doing this type of design work. There's a meteor called Memory Prices that are coming for consumer hardware and robotics and physical AI. We're in trouble as an industry. Today my guest is Caitlin [01:30] Caitlin is one of the most sought after and accomplished hardware leaders in Silicon Valley. [01:34] She was part of the original unibody MacBook Pro teams and technical lead on the MacBook Air and Mac Pro at Apple. [01:41] She led the AR Glasses hardware team at Meta, including the team behind Orion, [01:45] their most advanced AR product, [01:47] Before that, she ran the VR hardware team at Meta, where she helped design all of their incredible VR devices like the Rift and the Quest.

Most recently, she was at OpenAI, helping build their robotics and hardware division from scratch. Robots and hardware and physical AI are so hot right now. Every AI company and so many startups are launching building AI hardware products. And Caitlin has been at the center of this emerging field for decades. This conversation goes in a lot of different directions, many that I did not expect. [02:17] on the hardware side of building over the next few months. [02:20] Before we get into it, don't forget to check out Lenny's Product com for an incredible set of deals available exclusively to Lenny's newsletter subscribers.

With that, I bring you Caitlin Kalinowski. [02:32] Caitlin thank you so much for being here welcome to the podcast thank you so much for having me I'm excited to be here we're gonna go in a bunch of different directions I'm gonna bounce around [02:43] I want to talk about VR. So much money, so many resources, so many smart people have been working on VR for so long. [02:51] Meta spent, I don't know, $10 billion. Like they renamed the company Meta to lean into VR as the future of this metaverse that we're going to be living through.

[02:59] Feels like a lot of people are leaning out now. Feels like Meta stepping back, Apple stepping back with the Vision Pro. [03:04] In spite of the incredible hardware that everyone that you built, that your team built, just like I've got a couple of the devices. It's just like a magical experience that you unlike anything you've ever experienced. [03:15] still has not caught on [03:17] What happened? Is there still a future where VR catches on or is the future kind of AR and something else? [03:22] I don't think I would have guessed exactly what happened here, but [03:27] The way I look at it is [03:29] VR helped us understand how to orient things in space.

[03:33] relative to a simulated world and the real world and connect those two [03:37] We figured out Slam. [03:39] which was how to do positioning in space using cameras, [03:42] We figured out a lot of applications of depth sensors. [03:46] We figured out how humans perceive visual data, [03:51] in space, and all of that actually... [03:54] while it's great for VR and I think VR gaming's a really interesting, it is kind of a niche, but I think it's an interesting niche. [04:00] What I see now is in robotics, all of these technologies are being used.

[04:04] because you need to understand [04:06] how the robot is moving through space. You need to understand how far it is from everything. You need to understand [04:12] if you're wearing a VR headset and driving the robot, it's the same real technology. And so [04:17] For me, I view it as a [04:19] step in a long technological [04:22] arc [04:23] And to be honest, as someone who's not using VR a lot right now, [04:28] I'm really glad that we did it, but I don't think it I expected it to be big, obviously, or wouldn't have been working in Oculus.

[04:35] And [04:36] I... [04:37] think maybe the social aspect of having something in front of your face is [04:42] is part of why it didn't take off. And I think that we learned [04:45] of course, with Google Glass, how important that is as well. [04:48] And so when we tried to make it social, [04:51] Um, it's hard to make it social when you have, you know, your face covered. [04:55] That is interesting. So just like the investment in... [04:58] innovation that happened that went into VR has actually proven to be really useful.

[05:02] And so it feels like the companies that have put a lot of effort into that and money into that have... [05:06] are ahead on the next step. So is that where you think things go? What's kind of like, where do you think things are going? Is it AR glasses or something else? What's kind of the future of this? I believe in AR glasses as part of the future because... [05:19] I do think looking down at your phone all the time is not great for us as social creatures. So if you can [05:25] maintain social connections and get information.

[05:28] That's where I think we're headed. [05:32] Orion, the AR glasses we worked on, I worked on most recently, [05:36] are a bit ahead of their time because they're using [05:38] waveguides and micro LEDs that are not quite ready for mass production. The yields just aren't there. The cost is still high. [05:45] I think that's absolutely a path that air glasses are likely to take. [05:49] And as we figure out the input to those glasses, like how do you communicate with them when you're on the move, when you're in public?

[05:55] How do you communicate quietly, silently? [05:58] with them, I think, once we start to figure out some of those [06:01] challenges that [06:03] having a display that's mostly off. [06:06] that you can turn on when you want it to be on seems like part of the future. [06:10] So that's part of it. The other part is there's this lineage [06:13] of technology going through VR and then AR and now in [06:17] I'm using the term robotics, physical AI, but you really have to step back and look at [06:21] autonomous vehicles, drones, [06:24] Obviously robots, robots.

[06:26] autonomy period, manufacturing, like all of these technologies are going to need the same [06:30] the same piece parts, the same pieces that we built in the AR VR spectrum. [06:35] It's interesting with VR, there's this idea with when you build product, there's always this question when something doesn't work. [06:39] Is it just like you executed it badly or the idea was just a bad idea and it's always hard to know? It feels like with AR, it's just like... [06:46] so much effort was put into making it work just like for a decade, many decades.

[06:51] and just has not worked. So it's like nice that we know, okay, there's nothing we can really do right now to make this work. [06:57] I completely agree with you. The issue is just like, I don't want to sit on my couch. [07:00] disconnected from the world. And even if I could see people through it, it's just like, I'm just gonna, I don't need this. It's not that big of a deal. And AR, you're gonna just start getting more and more larger and larger displays. But the great thing about Orion is you got 70 degree field of view binocular.

So with the prototype, you got to sense what this is really going to be like in the future. [07:20] It's very hard to describe how it feels. [07:23] to use a pair of glasses like this but when you do you suddenly are like oh like i feel immersed it's the field of view is wide enough i feel immersed and it becomes pretty clear that i think [07:33] I think this is part of where the future is headed. [07:35] This episode is brought to you by our season's presenting sponsor, WorkOS. What do OpenAI, Anthropic, Cursor, Vercel, Replit, Sierra, Clay, and hundreds of other winning companies all have in common?

They are all powered by WorkOS. If you're building a product for the enterprise, you've felt the pain of integrating single sign-on, SCIM, RBAC, audit logs, and other features required by large companies. WorkOS turns those deal blockers into drop-in APIs with a modern developer platform built [08:05] specifically for B2B SaaS. Literally every startup that I'm an investor in that starts to expand up market ends up working with WorkOS. And that's because they are the best. Whether you are a seed stage startup trying to land your first enterprise customer or a unicorn expanding globally, WorkOS is the fastest path to becoming enterprise ready and unblocking growth.

It's essentially Stripe for enterprise features. Visit com to get started or just hit up their [08:35] WorkOS allows you to build faster with delightful APIs, comprehensive docs, and a smooth developer experience. Go to com to make your app enterprise-ready today. [08:45] Okay, I want to talk about robots, robotics. I was meeting with a bunch of Princeton students a couple months ago and they're [08:53] They're kind of like comp sci students, and they were telling me that [08:56] Enrollment in Comp Sci at Princeton is down, trending down. [09:00] And I confirm this is actually true at a lot of universities.

There's a lot of charts that show comp sci enrollment down. [09:05] and where it's actually going up is hardware, robotics, which [09:10] I imagine as someone that has been in this field for a long time is very weird because it's never been that popular. Just how does how does it feel to feel like a while everyone's getting in now? It's very odd. Everyone is suddenly asking about hardware and robots and the physical world. [09:25] And it's never been a sexy career. [09:27] It's always been the thing that you went into because you loved it.

It never paid the same as these other careers. It was never kind of at the forefront of how we talk about things, with the possible exception of Apple, obviously, and the hardware lineage at Apple. [09:41] So it's... [09:42] It's great in some ways and it's very odd in others. [09:46] What's surprisingly hard about [09:48] Hardware. A lot of software companies, a lot of people are just like, OK, cool, we're going to build some hardware. That's the future. That's the mode now. [09:55] And they get into him and they're like, what the heck?

[09:57] What are some things that maybe people don't think about when they think about, okay, we're going to build some hardware? What are some of the surprising challenges that come up? I like to talk to computer science folks about it this way. [10:06] So computer science folks, as you know, [10:08] They write code. [10:10] And then they compile the code off it and they run the code and debug it. [10:15] But they can compile their code every day, you know, every hour, whatever they need to do. [10:21] In hardware, we only get to compile our code, quote unquote, [10:24] like four or five times.

[10:27] Four or five times a year or... [10:29] So long. [10:30] Okay, ever. [10:31] Right. So if you're building hardware, [10:34] You redesign it in CAD. [10:37] you know for every major build [10:40] and then you have to release it and once it's released you compile it the last time you release it for mass production [10:47] If it's a mass production device, that's it. [10:49] you're done you can't ship over there updates so we have a different approach we have to have a different approach [10:56] which is more conservative. [10:57] You have to do more of the reliability checks and tests [11:01] in line with the program because once you compile that last time you're done [11:06] You make all the parts, you put them together, they're out in the world.

The only alternative is to [11:12] ship something new to replace it a couple years later. And so we have to be more conservative and we have to take our time [11:18] because we're [11:19] If you think about it, [11:21] a product that sells millions [11:23] If you have a graph of all the parts put together on any different part of the of the device, you have a curve. [11:30] you're in the plus and minus three sigma or more. [11:34] Right. So meaning [11:36] if you have two parts that go together, you're going to get the smallest version of this one and the largest version of this one, and you're going to have to put those together across the board.

[11:44] People don't think about this that much, but the part variance is pretty high. [11:48] And so we've got to solve for that last half a percent. [11:52] in the process of building, [11:54] so that when we compile our last time, when we build our last time, it's done. [11:59] And we're not going to have, we're going to have a high yield. We're going to be able to make [12:02] them and make money on them effectively, we won't have very many returns. [12:06] And so that's kind of the game that we're playing.

- It sounds so hard and complicated. They're just like, Safra is so nice. [12:12] Write some code, ship it. It's great. Why do you think people are getting so into robots and hardware now? What's kind of the driving trend? [12:20] Yeah, what I'm seeing in the [12:21] you know, in the AI world in San Francisco, [12:25] is there's a dawning realization, especially in the labs, I think, [12:29] that the acceleration is going so vertical [12:32] That way you can do behind a keyboard with AI is going to saturate. [12:36] Now, I don't know when it's going to saturate.

Nobody else knows either. [12:39] But when that happens, the next the next frontier is the physical world. [12:44] And so what I see happening is the labs, [12:48] Big tech. [12:49] startups are all realizing at the same time, okay, [12:54] This is coming. [12:55] We're going to have complex problems [12:58] systems that can solve problems in the digital world very, very quickly. We already have them. They're going to get better and more comprehensive and more capable. [13:07] If you think about that as a frontier, [13:10] you can see the end of that tunnel.

Now, I don't know when it's going to be again, but we can see that that's going to saturate at some point, or at least people think it will. [13:17] And when that happens, the next frontier is hardware, the next frontier is robotics. [13:21] manufacturing, industrialization, the sensing layer in the real world, [13:27] the ability to move objects in the real world, [13:31] and eventually, we hope, space. [13:33] So one of the most interesting lines of development is these humanoid robots. That's kind of like, you know, our meat brains are always more attracted to robots that look like us and act like us.

[13:43] There's a few companies very ahead. There's Optimus, Tesla, there's Figure, there's NIO, there's a few others. [13:50] What's your sense on just the current state of these humanoids and kind of [13:54] I don't know, like how close are we to human beings being around us? [13:58] We might be close. [13:59] I have [14:01] like many others, safety concerns about large, strong humanoids operating right next to people because [14:07] we have to have enough data to show that that's safe. [14:11] There are some designs, and 1x Neo is a good example of this, that have made

[14:16] significant safety considerations in their designs and pulled mass inwards essentially [14:22] which is a lot safer. Softer robots is safer. Just to clarify, you're saying they're lighter, and so the impact of a robot hitting you is less. [14:30] Yeah, the part that might hit you, which in this case might be the arm. Mm-hmm. [14:34] if it's lighter. [14:35] and softer. [14:36] There's two aspects. [14:38] You have... [14:39] The arm moving through space. [14:41] and then you have the actuator that's rotating. [14:44] So you have to add up the energy essentially for both of those things.

[14:48] Um, and [14:50] So that's an impact thing that you have to worry about. [14:53] Then you have to worry about the compliance of the arm. [14:55] If it's just... [14:56] Hard? [14:57] then the impulse is high. [15:01] But if it's soft and compressible, then the impulse is lower. [15:04] And so... [15:05] You really have to be thinking about this when you have robots around people. [15:08] So in my world, in my worldview, [15:11] the humanoid robots are still prototypes. [15:14] and their advanced prototypes. [15:17] What we need to do is show that this works at all.

[15:20] which is kind of where we're at right now. [15:22] Once we have working prototypes, then usually, at least in my field, what you do is you... [15:27] you [15:29] continue to revise them [15:31] to make them cheaper, [15:32] easier to manufacture, higher yield, and safer. [15:36] And I think this is what's going to happen next. So they're not quite in my in my mind, they're not quite ready yet. [15:41] I even get [15:42] you can get a Chinese robot that can do all kinds of things for you. But if you look at the booklet, it says, hey, you can't be within three feet.

No human can be within three feet of this robot. [15:51] And you're not going to see very many robots that are strong enough to do meaningful work. [15:56] that don't have that warning right now. [15:59] That is so interesting. It's funny to hear that at the same time there was these nunchuck-wielding robots in China doing dances with other folks. I've never thought about just that part of it, like the impact they can have if they go... [16:12] Array. [16:13] I want to come back to that, but just... [16:14] Like timeline wise, what's your sense realistically when...

[16:18] humanoid robots are walking around the streets in people's homes kind of at scale. [16:22] At scale, [16:23] is the problem. [16:24] in my mind at scale is a huge challenge. Now for me, my background at scale means millions usually. [16:32] But let's even say hundreds of thousands. [16:36] You've got to get a good design that's running. [16:38] then you've got to make it reliable enough that it can keep running day to day to day without a lot of human intervention or repair. [16:45] And that's its own problem. [16:47] But the first probably you have a supply chain.

[16:50] And this is going to be something that I hope that we can talk about a little bit more. [16:53] But every single part that goes into that robot is coming from somewhere. [16:57] And [16:58] Many of these parts may become more restricted or difficult to make. [17:03] And it may be harder to assemble [17:05] the subassemblies [17:07] and the meaningful parts of the robot here in this country. [17:10] So there's a very complex supply chain dependency right now. [17:14] on robots like humanoids, but also other robots that we have to [17:18] that we have to figure out.

And a lot of people are trying to [17:22] move production, [17:24] here to the united states [17:25] which is very challenging because we don't have [17:28] great actuator companies here yet. [17:30] for example. [17:31] And the actuator is like the little arm... [17:34] uh i don't know how would you describe an actuator to a non-robotic person yeah the actuator is the motor so you put power into it electricity into it and you get motion out of it and most of these [17:46] a rotating... [17:47] rotor essentially [17:48] that then has gearing on it that then powers the limb or powers the head or the fingers or whatever else.

So they can be small, they can be large. Okay, awesome. I hear the word a lot. I'm like, I don't know exactly what it means. Thank you for explaining it. [18:03] I want to talk about the supply chain stuff more because I know you think a lot about this. [18:06] What's kind of like the state of the union on the supply chain for say robotics? What's going on? What are the pieces? What's what are the challenges? [18:12] So the way to think about it is you can start with raw materials and magnets is a good place to start.

So we need to be able to get the magnets, the raw magnets. [18:20] For example, [18:21] Then we need to be able to process them. [18:23] Then we need to be able to integrate them into actuators and build the actuators around them. [18:28] that we need to be able to integrate those actuators into subcomponents or robots themselves. [18:33] And each layer of this chain [18:35] has essentially been outsourced over the last 25 years to countries like China, [18:40] like Japan, like Korea. [18:42] And so, and full transparency, I've been part of that transfer of engineering knowledge.

[18:50] to Asia. [18:51] in Asia, the expertise has historically been [18:54] scale. [18:55] and being able to build a lot of these parts at lower prices. [18:59] We've had this kind of deal. [19:01] across these borders that this is how we're going to operate for the most part now of course [19:04] There's things we make in this country still. [19:07] And of course, there's design design. [19:09] and AI that's made in Asia, but that's essentially where [19:12] things have been for a long time. [19:15] And in order to [19:17] have a safe supply chain, we needed to start to work on [19:21] having some independence in these layers and these stacks.

[19:24] And it's interesting that your focus is on these like actuator like that. Is that the bottleneck, this very specific part of a robot? [19:30] It might be. It might. So if we can't get the magnets, then we have to design new actuator types that are maybe use different materials that may be larger, that may not be as efficient in space. [19:41] So that's important. And then the actuators themselves are important because if for some reason we can't buy them, [19:46] then we don't get to make robots. So it's foundational.

There are some foundational technologies like this. [19:52] all backed by material science, essentially breakthroughs. There's batteries, of course, [19:57] There's actuators. [19:59] the raw parts [20:01] like the die cast parts, um, [20:03] the machine parts are less... [20:05] critical, we think we can get those. [20:07] But we're [20:09] I think everyone... [20:11] not just in this country, but around the world, is starting to think about supply chain. [20:14] because you have these disruptions, whether it's COVID or war, [20:18] and you see how quickly things change. - Okay, stupid question. Why magnets?

Why is that a part of the supply chain? Why do we need magnets? - Yeah, so it's a great question. So you have a ring of magnets. [20:28] that are polar opposites and they go like this around the ring. [20:32] And then... [20:33] and then you have you know you have something in the center that that rotates and the way it rotates is you have alternating current [20:41] essentially. And so the magnets make the [20:44] the rotor spin. [20:45] Wow. We need to have a YouTube lecture of here's how this physics works.

Okay, very cool. [20:51] So when you talk about China, this is like what I imagine, what I think about now is watching the war in Ukraine and Russia, just the drones. [20:58] just like how crazy and different the world is now that you can build these little drones that go and you know blow people up robots are part of that it's just like such a [21:08] existential threat. [21:09] to every country now. [21:11] the ability to build these things at scale. [21:14] What's your advice? What should we do? What should we change to be, you know, to thrive in this future and not be, you know.

[21:20] in trouble. [21:20] Well, you mentioned drones. It's another good example. You need essentially the same technology to make the rotor spin on a drone. [21:27] as you do to make [21:28] an arm move on a robot. It's essentially the same base [21:31] technology. [21:32] and supply chain. [21:34] So we need to, at least on the military side, have an independent supply chain as much as possible. [21:38] I think that's important. [21:40] I think every other country should do that as well. [21:42] but I don't think that's specific to us.

[21:44] Um, [21:45] I do feel that we need to re-industrialize the country significantly. [21:50] in order to be safe. [21:51] in a military sense. [21:53] You really never know what's going to happen in the future. And the people that are your allies now [21:57] may not be in the future. [21:59] the Allied West, I think, is [22:02] is going through a lot of geopolitical changes, [22:05] There's a lot of shifting. [22:07] And so I would really like to [22:11] reteach ourselves how to make things at scale, how to make things at quantity.

[22:15] how to, [22:16] process raw materials. [22:18] how to be more [22:19] um independent so that when covet happens again or something else happens again [22:24] We're not in trouble and we can't. [22:25] But [22:26] and we're not unable to [22:28] you know, protect ourselves. [22:29] What I think about also is Marc Andreessen had this visual on some podcasts of just imagine 100,000 drones just coming out of China just at us. [22:37] What do we do? We're not prepared for that. I don't want to spend all our time on this dark stuff.

[22:41] But it's a real thing. [22:42] Well, and Palmer Luckey is a friend of mine. [22:46] And we don't agree on everything, but I do think that we agree on some important aspects of how we need to respond here. [22:55] I think he's right to say that we need to invest a lot more in drones. [23:00] than in aircraft carriers. I think that is this old way [23:04] of thinking. [23:06] And these are important components of the military. [23:08] But it's an old way of thinking of, hey, we have this and we have this and we have this and our planes come off here.

It's like, [23:14] No, AI is changing everything. [23:17] And, um, [23:18] Military technology is changing incredibly fast and the place to look at that is Ukraine where [23:23] you know, drones are being changed and updated every day rapidly with 3D printing. [23:28] And this is, I think, the future of where war is headed. [23:32] unfortunately, and I view this as a very different era that we're entering into with very different [23:38] It's a, you know, this isn't new to anybody. [23:41] But this is a, you're looking at what [23:43] It costs for them [23:44] to send out a missile and what it costs for us to stop it.

And this is a just you have to do the math every time. And right now we're losing on the math. [23:52] which is fine [23:53] for a certain amount of time, but [23:55] the longer it goes, the less fine it is. Are you optimistic that we'll figure this out? [23:58] Yeah, America is really good at figuring these things out. [24:02] that we have a pioneering kind of [24:04] independent spirit and a great engineering culture um but we need to we need to move [24:10] It's interesting that we started the conversation with VR.

Palmer Luckey obviously famously started Oculus. It's interesting how this is so connected. You think VR is this trivial thing that we're just playing games and such, but it's like the same person is now building Anduril, which is... [24:26] the leading, I don't know, for robot building hardware company. Yeah, and I think we need a lot more of them. You know, I've chosen not to work for companies that [24:36] create lethal technology, um, [24:39] And but but I think that it's good to have people who are willing to do that. [24:43] And I think that it takes everyone kind of to build [24:46] the future that we want.

[24:47] Mm-hmm. [24:48] Coming back to the AI safety piece, it's so interesting. I had a couple of conversations like this on the podcast recently. [24:53] We think about all this like prompt injection and jailbreaking that happens with chatbots. [24:59] And we. [25:00] Like, not enough people think about what if you prompt inject a robot walking around and tell them to punch someone. [25:05] And we're like so far from that feeling like we can actually stop that. [25:08] Yeah, we have to be able to control [25:11] adversarial threats to our hardware lair [25:13] whether it's robotics or drones or anything else.

And that's going to be a huge part of the future of warfare. [25:18] Yeah, just like people talking about OpenClaw and how much you could just tell it. You know, there's all these, like, give me all your passwords. And it's done all these things to people's lives and just like... [25:28] Robots walking around, hey, okay, here's all this person's secrets. [25:33] My open class story is I have I sandboxed it. So it's on its own computer. [25:38] But I gave it like three things. I gave it like my real email address and and my [25:42] I don't know what it was.

I gave it like [25:44] some information about one of my accounts or something like that. [25:48] And I added it to the social media, like, I can't remember what it's called, the open claw. Oh, Maltbook? Yeah, I added it to Maltbook, and I was like, okay, whatever you do, don't share my private information. But, oh, crazy. And five minutes later, all it had done is posted my personal email address. It was like the one thing it had. Nailed it. Okay, you're shut down. It was so funny, no matter how careful you are with these things.

[26:11] You just can't really, we're not at a place where I think. Which is exactly your point, that the robots can do a lot more damage. And I never thought about just like the softness of their hand as a way to keep us safer. [26:22] Yeah. [26:23] Oh, man. And Nat Friedman just did this interesting talk at Stripe Sessions, and he was talking about he's talking to his open claw about drinking more water. [26:30] and sleeping better and and it has these driving in a self-driving car [26:34] It told them, OK, here, there's a place off the freeway that you should go to.

And it changed. [26:39] the destination of his Tesla. [26:41] to take him there because I imagine he connected it to their API at some point. [26:45] That's so funny. Oh, my God. Yeah, these are going to get weird fast, I think. [26:49] Okay, so kind of on this thread of... [26:53] hardware... [26:54] emerging as a moat, as something people realize is a big part of the future to be competitive, AI labs, all these other companies. [27:01] You've been at [27:03] a company's you've been at Apple, which had a very [27:06] great and long-lasting hardware program.

[27:09] Then you went to Meta where you helped build, basically bootstrap a hardware program from scratch. [27:15] I feel like those lessons are very valuable to people trying to do that now. [27:19] What was the experience like helping Meta build a hardware program? And what are some lessons for people that are trying to do this at their company? So Apple has been best in class at this. [27:28] Um, [27:29] There's a bunch of reasons. One, hardware is a first tier citizen at Apple. [27:34] There's a lot of companies where hardware isn't part of the core...

[27:37] product development conversation as much, but that's an exception. [27:42] Apple also taught me [27:44] And a lot of other people, actually, if you look at kind of the era that I was there, I was very, very lucky because. [27:49] If you look at the other folks who were there, [27:51] I was there between 07 and the end of 2012. If you look at the other people who were there working on these things, [27:57] They actually have a lot of key positions now across the industry. [28:01] And I attribute that to how [28:04] good apple is at training people [28:07] to think.

[28:08] about complex interdependent [28:11] decisions and risk. [28:13] And I don't think I realized that they were doing that at the time. But if you look back, what you see is a real dedication to hardware excellence. [28:23] the proper process to go through and do really good experiments in hardware and figure out what the best outcome is. [28:32] But there's something underneath that which is understanding the first principles of why are we building it this way? [28:37] and what are the key outcomes we're looking for. [28:40] And actually, John Ternus talked about this, I think, [28:43] a few days ago where he talked about the back of the cabinet.

I don't know if you saw this video, but basically John said, [28:50] that he was impressed that he learns from Steve Jobs that there's a cabinet maker who finished the back of the cabinet and how important that was. [28:57] And that goes very, very deep. [28:59] at Apple where every single design decision, even on the inside of the device, is considered. [29:05] And this isn't just an aesthetic concept. [29:08] decision. [29:09] What it does is actually force [29:11] the engineering [29:12] industrial design operations community there to think about what are we really doing and what's the core of what's happening [29:19] for this part.

[29:20] for this assembly, for this consumer product, [29:23] then what really matters and what happens is if you're if you're that methodical, [29:27] what really matters tends to rise out. [29:29] and look very simple at the end. [29:31] And so part of what you're seeing in... [29:34] Many folks coming from that era [29:36] is an understanding of how to do that. [29:39] which... [29:40] you know, in the very beginning of the Mac side, Mac's [29:44] Didn't sell as many. [29:46] and the quality wasn't quite as high. [29:48] But by the end of that era...

[29:50] you know, Macs were very popular and selling in much higher volumes. [29:55] And so I think that made a big difference. And I was only a small part of that. Like I was, you know, the thermal lead on the first MacBook Pro. [30:03] and then over time worked to lead successive [30:06] iterations of the MacBook Air and the cylindrical Mac Pro [30:10] But I was lucky enough to work with these folks and learn from them who've been doing this for a really long time. So you have to take those lessons carefully.

[30:17] And then when you leave, [30:18] try to distill them and explain them to a new community. [30:22] Now, [30:23] Oculus was actually a hacking hardware startup. [30:27] Oculus started [30:29] from folks who actually met on forums. [30:32] You might know this, Lenny. [30:34] who were hacking... [30:36] like PlayStations or Super Nintendos into portable backpacks. So and and so there was an ethos at the company that was actually quite [30:44] good for the dna of a hardware team [30:47] And then I was on the meta side when we did the acquisition.

And when we acquired them, [30:52] They had that spirit of rapid iteration. We they had made Crescent Bay [30:57] before the acquisition, I think. [30:59] But then to professionalize that [31:03] get the yields up and get the volumes up was was [31:07] the cost down. [31:09] was kind of the challenge we faced in the first Rift. [31:11] So one lesson I'm hearing here is being very... [31:15] detail-oriented, I don't know if that's the right word, just like focus on every element of [31:19] of the end product because to your point, it's not just about that back of the cabinet, but it's like I think about it's like the brand M&M story where like a band puts in the contract, you have to have brand M&Ms in them.

[31:30] in the room because that means they read it and it's not like m&ms matter it's like it's a test that they read the thing [31:36] And is that kind of the message there? [31:39] I think the message is understanding why you're doing what you're doing. [31:43] And then every design decision supporting that. [31:47] Goal. [31:48] And that requires a lot of detail and it requires a lot of persistence and that requires a lot of consistency. [31:55] But understanding why you're doing what you're doing and what the end goal is, is, is, I think, the key.

[32:00] and letting that expand into not only the software and the UX, but also the hardware. [32:05] What's an example of that, just to make it more concrete for us? [32:08] A great example is the Quest 2. [32:12] So we reduced the quest to price quite a lot. [32:16] And what we had to do is understand what is what are we trying to do? We're trying to democratize VR. We're trying to get VR to more people. [32:22] And the only way we could do that is reduce the price. [32:25] And so what it required is a redesign.

[32:29] of the entire product essentially for cost. [32:31] which [32:32] then I think led to the highest selling VR headset of all time. [32:37] And it's not easy because you had to, in our case, [32:40] remove cameras remove components [32:43] change materials, change manufacturing processes. But when you have alignment that you want to get this to more people, and the way to do that is to reduce the cost, [32:51] then that kind of drives everything else. And it was still a very high quality product with with [32:56] with great, I think, low return rates, and it was a very strong product.

[33:01] Um, [33:02] maybe even stronger than if we hadn't done that, funny enough. [33:05] But it hit our hit our price point. [33:07] Okay, coming back to just the question of, say companies like, okay, we need to build some hardware, we're going to build our own glasses, we're going to build a little phone device and secretive thing, whatever opening is up to. [33:17] Uh, [33:17] What other... [33:18] tips do you have? I know it's like impossible to like, here's all you need to know. But just what else? What else should people be thinking?

Having your goals defined early and sticking to them is important. [33:29] Hardware is not as... [33:32] adaptable to lots of changes throughout its development as [33:37] anything digital. [33:39] And so if you set out to say, okay, we want to make something that's $300, [33:43] And then halfway through you say, oh, it actually has to be $150. [33:47] You've almost burned a lot of that early time. [33:49] So you kind of need to have a sense of having pre-thought out what you want. [33:53] and having those, I like to call them KPIs, but essentially goals.

[33:57] written down and try to change them as little as possible. So that is very tough. In fact, that may be the toughest thing. [34:04] because... [34:05] If you do that properly and you and you have, you know, [34:09] the right prioritization of those things. [34:11] you know whether you can ship or not. [34:14] you know whether you're done. [34:15] And in hardware, one of the challenges is, [34:18] You know, we talked about compiling four or five times. [34:20] every time you build, [34:23] and you iterate your design, [34:24] that's another three months or four months or five months or whatever it might be.

[34:28] And so you're trying to time [34:31] the feature set with the quality, with the timing. [34:34] And in hardware, timing is important because if you come out with your product a few weeks before your competitor, [34:39] You might get all the PR. [34:41] You might get all the interest. [34:43] It's pretty brutal. [34:44] And so [34:45] Each of those days that you ship before your competitor is worth a lot of money. It might be worth $10 million to you. I'm making this up. I don't know. [34:52] So you have to balance that with how many times you iterate.

[34:56] And if you know what your goals are up front and you hit them, then you know you can ship. [35:00] And often engineers, and I'm guilty of this too, especially on the hardware side, never feel like they're done. [35:06] So this is a pretty nuanced thing. So that's one thing. [35:10] The second thing is we tend to design the things... [35:13] that we know how to design first. [35:14] And actually the right approach is to design the hardest parts first. [35:18] One example will be, and there's no IP here, so I'm obviously not going to share any.

[35:23] any any IP or anything internal. But at one point, we had to route cables through a hinge. [35:28] in a device, in a laptop we were making. [35:31] And because it wasn't clear that [35:33] those cables would fit. That's where the architect started. [35:36] And he looked at the cross, the diameter, and how to split the cables out, and made sure that they would fit before finalizing the hinge design. [35:44] A lot of people would start [35:46] at the part they knew, like, oh, we're going to use this display.

So I'm going to put this in CAD and I'm doing this stuff. But the architects who [35:51] who are the best, actually look at. [35:53] where are the pinch points where is this going to fail and they start to do the detailed design there first [35:58] And then a couple other points is... [36:00] the part that you're [36:02] customer touches or interacts with the most. [36:04] needs way more iteration than everything else. So [36:08] easy on a computer, you touch the trackpad the most, and then maybe the [36:12] the keyboard next.

So those things have to be really good. They have to feel good. [36:15] They have to respond properly. They have to be highly reliable. [36:19] And then maybe the other pieces further out [36:21] don't take quite as much iteration. [36:24] So you have to boost your iteration on the things that [36:27] people touch the most or interact with the most. [36:30] Um, [36:30] So those are kind of some principles that I wrote about. [36:33] But these are just things that you learn. [36:36] trying to build quickly. [36:39] And the last piece that's really critical for making hardware [36:42] for folks out there who are trying to make hardware is [36:45] You can't wait around ever.

[36:47] like there's never enough time. [36:50] So if you know that you need to do something, what I learned from [36:54] from folks like Shelly Goldberg at Apple now, who I think is a VP now. [36:58] and Kate Bergeron. [37:00] when I was there at Apple is, [37:02] You need to do it right now. Anything you know you need to do, you need to do right now, because in two days, there's going to be a surprise coming around the corner that you need that time to fix. [37:10] And so this sense of stacking [37:12] the things that you know you need to do, [37:14] and just getting them out of the way.

[37:16] even if you technically have more time. [37:19] is this like kind of ruthless efficiency that I learned. [37:23] with them. [37:24] Amazing. Okay, let me just summarize your advice here. So one is [37:27] Be very clear on goals. [37:29] I want to come back to this. [37:30] Two is do the hardest part first, the riskiest piece, essentially. [37:34] to physically build. [37:36] Three is focus on the [37:38] pieces that people will use most, say the trackpad. [37:41] uh, [37:42] keyboard. I want to talk about that. And four is just like, do it now.

Even if you think you have more time, this is going to [37:48] You never know what's around the corner. You just don't. [37:51] It's not even that you don't know what's around the corner. If you're working in hardware, like, you actually don't have more time. [37:56] Thank you. [37:57] Okay, on the goals, what are kind of like buckets of goals? So cost is when you shared like we need this under $300. What are some other like? [38:04] buckets of types of goals people should be thinking about. [38:06] So in VR, [38:08] uh display resolution or arc minutes um like how many pixels per degree do you want is actually one of the key metrics so you need to understand what your key metrics are [38:17] And why is that key?

Well, [38:19] That's your visual field. So you think about retina displays on MacBooks, [38:24] They figured out the KPI of what the human eye could see. [38:28] probably overshot it a little bit and built that and then do you really need to [38:32] keep as much engineering pressure up [38:34] on [38:35] the resolution of a display after that, maybe not. [38:38] So VR is not there yet, not even close. So not in mass produced VR. We don't have retina displays yet. So that is [38:46] one aspect of pushing that up is one example.

[38:48] I think on a computer, obviously, you're talking about [38:51] Clock speed. [38:52] You're talking about how many parallel processes you can run. You're talking about [38:55] Wait. [38:56] You're talking about price. [38:58] And you're talking about features. [38:59] So when we did the MacBook Air, [39:03] it became very clear because we were machining it that there are certain features like ambient light sensor that we just didn't make sense anymore. [39:10] And so being willing to just jettison them. [39:12] for what we were going for, which was weight. [39:16] and size.

[39:17] So if you have those overarching... [39:20] goals you can actually make decisions engineering decisions pretty quickly and this is actually something that i think elon [39:25] I've heard does very well. [39:28] is [39:29] define [39:30] the value [39:32] of... [39:33] you know a gram of weight versus [39:36] the cost or he does, I've heard engineering ratios essentially. [39:43] and he's able to put numbers on what those ratios should be, which I think is really smart. [39:47] Interesting. So it's a very easy trade-off. Okay, here's the [39:50] Here's the formula telling us weight is less important in this case.

[39:53] Yeah, and if you can do that, then the decisions fall out pretty easily. [39:58] Speaking of the air and weight, I remember... [40:00] I feel like there's a very classic moment in Steve Jobs lore where he comes out and has this manila envelope and has the... [40:06] MacBook Air inside it and then takes it out and they were like, no way. [40:10] Were you part of that or was that something that people wanted to do from the beginning? [40:14] I think if my memory serves... [40:17] The very, very, very first MacBook Air, [40:21] was a pretty low volume device.

[40:23] that was machined [40:26] but kind of had a proof, more of a proof of what could be done. [40:29] And that was a manila envelope one, I think, where the side door opened out to give you the port and it kind of had a... [40:36] It had this shape underneath. [40:38] And then... [40:40] The next rev of that was the MacBook Air that we know, which was essentially, which is wedge shaped, which is different. And so the wedge shape is the one that I worked on and the one that went and hit more volume.

[40:52] But that Manila envelope one was the one that proved you can CNC a computer. [40:57] And so they each have really important roles in the roadmap. [41:01] Coming back to your... [41:03] point about focusing on things that people use the most. Famously, Apple screwed up this keyboard. There was this butterfly keyboard situation for a long time. You're like, it's your clothes. [41:14] What happened, Caitlin? I didn't work directly on that keyboard. [41:19] Um, [41:20] So I can't talk about what happened with it. But obviously this is something that you gotta get right.

And I will say like the modern MacBook keyboards are awesome. [41:29] and excellent. [41:31] And, you know, I don't know what happened with that. I don't think those were devices I was working on at the time. Nice. Safe. Marked safe. [41:42] Along these lines, Apple is kind of famous for not [41:45] not listening to what people want. This is kind of like a classic thing with Steve Jobs. He's not walking around doing user focus groups, asking, doing user research. Somehow, [41:53] continues to build incredibly popular products. [41:56] What do you think they do, right?

Or do they do a lot of user feedback sessions, things like that? How does it end up working out? [42:03] It's been a long time. I mean, I left... [42:05] or a decade ago. I don't know what they're doing now in terms of user feedback. [42:10] I think this one gets misinterpreted, though, Lenny. I think that what is being said is... [42:15] is if you want to build something new, [42:19] Customers don't know what they want because they haven't seen it. [42:23] So a good example is the iPhone. [42:26] which I didn't work on.

But [42:28] When you build a new iPhone with a touch screen, you can't really go ask 100 people what they want because they're going to say a keyboard on their screen. [42:35] And this is, I think, the ethos that you're getting at, which is, and this is true for anybody building new product with a new feature. [42:41] And I've tried to build as much as I can teams that work on products that have something new about them. [42:47] Either they're a new category, [42:49] or there's a new manufacturing process or something that hasn't been done before.

[42:54] And when you're thinking about this, you can't really use... [42:57] what you learned from the same field and the same product class. Like it just doesn't work because you actually won't get the answer right. [43:04] And I think this is actually what [43:06] Steve was talking about, which is [43:10] You can't get intuition if you're changing something fundamentally, like your customers won't know what they want because they haven't seen it. [43:16] But if you show it to them, [43:18] They will absolutely know that it's awesome and that is what they want.

[43:21] But... [43:22] if you get stuck in an iterative feedback cycle with your customers, [43:28] it's very hard to go zero to one with something new. [43:30] And so in my view, I don't know for sure. I didn't talk to him about this, but that's my view of what that means. [43:37] I am so excited to tell you about this season's supporting sponsor Vanta. Vanta helps over 15,000 companies like Cursor, Ramp, Duolingo, Snowflake, and Atlassian earn and prove trust with their customers. Teams are building and shipping products faster than ever thanks to AI.

But as a result, the amount of risk being introduced into your product and your business is higher than it's ever been. [44:07] their organization, their business, and not to mention their customer data. Because things are moving so fast, they are constantly reacting, having to guess at priorities, and having to make do with outdated solutions. [44:19] Vanta automates compliance and risk management with over 35 security and privacy frameworks, including SOC 2, ISO 27001, and HIPAA. This helps companies get compliant fast and stay compliant. More than ever before, trust has the power to make or break your business.

[44:36] Learn more at com slash Lenny. And as a listener of this podcast, you get $1,000 off Vanta. [44:43] That's com slash Lenny. [44:46] I'm going to go in a completely different direction. Coming back to the components of hardware, I asked a bunch of people what to talk to you about. One of the people is the founder of Matic, the CEO of Matic. [44:57] Mahul Nari, Nari Awala. I've never said his last name out loud, so I hope I didn't butcher it. [45:02] By the way, I love my Matic. I don't know if you have a Matic, but it's like...

I have two, and I've purchased two more for... Oh, yeah. [45:11] What an endorsement. Yeah, basically it's like this amazing robot vacuum. [45:15] That just works. So his question, so he wanted to ask you, and so he suggests to ask you this, is about memory prices. [45:22] The way he described it is there's a meteor called memory prices that are coming for consumer hardware and robotics and physical AI. [45:28] What's going on there? Yeah, we're in trouble. [45:32] as an industry [45:35] I think that and I'm not an expert on this, but I think that AI has to do with why.

[45:40] And I also think that the supply chain is constrained. [45:45] I have been advising startups and companies to pre-buy memory. [45:49] and to have [45:50] enough memory in stock if they can afford it to ride out price spikes. [45:57] um like anything in this category uh [46:02] Let's see. This happened in COVID, too. Okay, so, like, we had... [46:06] so many supply chain disruptions and and getting enough memory was was one of the challenges so we had to pre-buy as well [46:13] I won't say who, but the company I was working at had a pre-byte memory as well.

And so this is – [46:20] part of what I wanted to talk to you about today is these supply chain disruptions and [46:24] If a key component that goes into a lot of tech like memory, [46:28] or silicon is constrained [46:30] There's not much you can do. [46:32] either pay [46:33] or you have already pre-bought enough [46:35] that you can ride things out. And so those are the only real options. [46:39] Obviously, there's a risk to pre-buying and the price might go down. [46:43] And so the challenge is I think there's a latency with supply chain in something like memory where it can't adapt fast enough often to demand.

[46:51] or there's a new category of product, [46:53] or in this case, maybe data centers that are just eating up so much and are actually not as cost sensitive as somebody in consumer electronics like Matic might be. [47:03] And so they'll just pay for these higher costs. This is tricky. [47:07] and something we have to deal with all the time. [47:09] How much of price has gone up? Like how bad is this problem? And then where do you think it'll go? Actually, this is a great question, Lenny. I don't know what's going to happen.

[47:17] I think prices are going to double probably. [47:19] um i don't know on what timeline if i knew what timeline the prices were gonna double on i'd be trading which i'm not very good like i'd really be i'd do be doing a different job if i could predict these things [47:31] But certainly we're not with supply chain shock. [47:33] And it's already gone up a lot. Like if you're saying it'll double, but it's already gone up. I don't know. I saw numbers like 6X. [47:38] Oh, really? [47:40] I didn't realize it was that bad.

That's a number I saw. It's not cool. I quote it. And, uh... [47:45] And you're saying, yeah, I think it from what I hear, it's AI driven, just like you need. And when you talk about memories like DRAM and things, what is memory when we talk about memory? What's going on there? Processing. It's the way to think about is like processing memory. [47:57] So it moves very you're able to kind of [48:02] you think about memory like on your hard drive or your solid state drive where you're keeping files that you're not using essentially in many cases or that you're

[48:09] you're dealing with, um, [48:11] you know, maybe documents or pictures that you have. [48:14] Maybe that's in cold storage on a server. Maybe that's using a hard drive somewhere. [48:18] This is usually things that you don't need really, really fast access on. [48:21] But if you're running a program, [48:24] some of that program is actually going to be run... [48:26] in RAM. [48:28] And so there's different kinds, obviously, for... [48:32] servers [48:33] There's different kinds of server racks. [48:35] Some server racks are actually focused on [48:39] this type of of of memory and some server racks are focused more on what we consider like a cold storage or a slower [48:47] Now, this isn't my area of expertise, but...

[48:50] Certainly most of the products that I've built [48:53] Maybe all of them have had RAM and we've had to figure out how to. For me, mostly it's a packaging issue. Where do you put it? Does it need to be accessible? [49:01] um uh you know which ram do you pick [49:05] How fast does it need to be? And what is the cost? Because there's usually our trade offs. [49:10] And what is the bottleneck with more RAM? Is it just the companies that make memory are just not able to produce at this rate because there's so much demand?

[49:17] That's right. That's exactly what's happened. [49:19] So this is a really good specific example of just how hard it is to build hardware. So this is just like all it takes is one piece to be not available and your whole thing is screwed. Yeah, you can't build anything if you have one component missing. So let's say Matic is an example. How many components are there that they all have to assemble and not have one not available? [49:38] I'm doing the math in my head.