Relentless
00:06:50 · Sam D'Amico
What's interesting is it's kind of the same thing that you mentioned, which is like the transition of China from like a low end manufacturer to a high end advanced manufacturer. I saw this at Oculus. So like Oculus was... The Oculus dev kits were built by, I believe, the same contract manufacturer that made Guitar Hero and Rock Band. And it was a toy... It's a toy factory. And so in some sense, it's like, they can do injection molding of, like, let's call it cosmetically, sort of sort of... I I don't wanna rag on them because I think it's like, you know, it's a different era, basically. Basically think like damp factory in Guangdong... Guangdong province, looked straight out of like the eighties sort of thing like that. It's probably a pretty similar site to what you would see if you were like going to China in the nineties or something like that. And then the advanced electronics piece was kind of kept into a like, okay, they put a computer on the factory line that like does some of the advanced, like software defined hardware stuff that like you need to do for a VR headset or whatever. So that was where they started when they got acquired. Then as soon as, as soon as the acquisition went through, I was involved with some of the initial contract manufacturer selection because I had worked on the Google Glass team on some of, the inside, you know, factory stuff there. And so got to see kind of the next wave of Chinese manufacturers and these folks are not mucking around. Like, this is like, it's not just that you get a factory, a building with some operators that they go and recruit, it's you get a full engineering team that co designs the product for you. And they co design the product around their supply chain, around their, like, like, around the tooling they've got, etcetera. So it's not just that you get a room with some cheap workers, it's you get a machine that effectively can build it for you. I I think the analogy for kind of the audience that that you've got is, like, it's kind of funny to use the Claude code for X sort of joke, but like, it's kind of like you have a LLM that can build hardware and you communicate to it through Google Slides and in person meetings and sometimes drinking some Baijiu at dinner. But like, that is how it actually works. And you were essentially able to be the high level engineering input to that system.
00:09:00 · Sam D'Amico
Now, there's kind of the split brain because you do that, but then you're also doing advanced R and D for like, hey, how does my computer vision system work? How, like, a lot of that stuff ends up has stayed stateside and like, you kind of see where, why this is like, you know, this is the, this is the same transition that you've seen kind of from like The United States still does the high value stuff. They still do the industrial design. They still do like, like Jony Ive didn't lose his job yet. Know, it's kind of the point like, and that team at Apple for instance has like one of every potential machine that they wanna use and they, they piloted different processes and they do it internally. But like, who scales it? Not them, obviously. And so I got to see kind of this transition from like, okay, China's this like cheap labor toy factory. You know, there's lead paint in the, you know, whatever, you know, all the, all the sudden people dismiss China with to like, oh, this has fundamentally changed. And like in the past, you know, couple years as I've been building Impulse, I've seen a different thing, which is like, these guys are not just really good at building stuff and like you can you can scale like an arbitrary VR headset or product through them. It's like they are leaders on the R and D side as well. And if you want the best battery cells, if you want the best process engineering for like how you build a battery pack, if you want like all this stuff, it's like, it's an email away from you getting it. And not only that, they are super eager to go and work with you. And they're super eager to go and work with you for a bunch of reasons. And a lot of them are, I would argue, fairly, like, fairly well intentioned. It's like it's like, well, they see, you know, The United States as, like, the benchmark of, you know, especially Bay Area engineers and stuff like that is like, this... We can learn a ton of stuff from this. And this will mean that we're ahead of our crosstown rivals in a hyper competitive, like like, you could argue that it's like all of these companies have been pumped up by state subsidies. And it's like it's like in China, there are 20 Elons. It's probably the way to think about it. In The US, there's one and the nimbies have basically tried to make it so that that
00:11:06 · Ty Morse, Sam D'Amico, Aaron Slodov
[00:11:06] S2: There's none. [00:11:07] S0: There's there's none. And and, but in China, they're they're they're they're they're like 20 to 50. Maybe that there... Maybe their Elon's are not as good. You can debate that sort of thing, but like that... That's how their hardware ecosystem and software ecosystem as well works. And the benefit of that is if I'm a American entrepreneur that wants to build the most advanced device of its category, I can just go there and do it. [00:11:34] S2: Yep. So, Aaron, I think Elon on a podcast recently basically explained that right now most of The US's biggest companies effectively just export files that they send to China, and then those files are converted into actual products. [00:11:46] S0: Mhmm. [00:11:47] S2: With this, like, massive decline over the last forty five years, how do you actually think The US can reverse this decline? [00:11:53] S1: So part of, you know, this story and building my company and seeing this stuff kind of firsthand led me to also, you know, try to start a national movement around this, which is re industrialized. Right? And the idea of that is we have only so much, you know, tribal knowledge left in terms of how these systems work, how you can ultimately, you know, create a machine like that. But once it's gone, it's it's gone. And preserving tacit knowledge is very hard because it's not like the matrix where you just jump in and download. [00:12:35] S0: I know kung [00:12:36] S1: fu. Right.
00:12:36 · Sam D'Amico
[00:12:36] S0: Yeah. No. There's a take on this that's like machine... So I would also argue that like, effectively you machine learn or you reinforcement learn how to manufacture stuff. And China is incredibly good at this process. And so basically it's like, okay, if I'm... We can break this down to like a simple example of like, let's build a lithium ion battery cell. It's not really simple, but there's all of these process steps where you're basically like laminating film, like laminating films on top of each other with different chemical processes and like the like the makeup of the the powders you're using and like the settings on the machines that do certain steps. All of that stuff you can imagine, again, I'm I'm assuming your audience is full of a bunch of aspiring AI engineers or whatever. All those things are like implicit weights into a neural, like a neural network basically. Yeah. And you can pre train, you know, you can take all the knowledge in the world and design the best thing, the best single widget, but there is a reinforcement learning with factory feedback step that unless you build stuff at high volume, you will never achieve. And this is actually the thing that like, I've been talking to a number of folks doing kind of like new manufacturing things. And it's like, that is the key thing that I think a lot of people who have not been close to the process, they don't get basically. Yeah. Is this like implicit reinforcement learning aspect of it and the fact that like so much of the state of what makes advanced manufacturing work is not in the spec.
00:14:01 · Aaron Slodov, Sam D'Amico, Ty Morse
[00:14:01] S1: Yeah. It's it's a... If you think about it like a stateful system, it's impossible to observe until you have high throughput, which is mass production, basically. So... [00:14:11] S0: And who's gonna fund you for that? [00:14:12] S1: Exactly. It's it's very hard to understand all the parameters basically until you do it. So that's why, you know, the Chinese machine is so efficient because they are just gathering more and more tribal knowledge and it's gonna snowball out of control, right, until basically, like, we'll get to this later, but the humanoid thing is an insane problem, if it actually plays out. [00:14:38] S2: What do you mean by that?
00:14:39 · Aaron Slodov
[00:14:39] S1: Well, we'll get there. We'll get there. But my like, basically, the whole the whole purpose and MO of re industrialized, right, is that, I would prefer if we could preserve as much of that knowledge as we can, encapsulate it, and then rebuild it here. Right? And be able to have these kind of like reinforcement feedback loops because we're building better brains and like compute. And it's, in my opinion, it's not necessarily that like we need to live in a unipolar world with a single industrial superpower, but like there has to be another option. And I also... I don't believe that, like, autarky is a necessity. Right? Like, you can't... Trade is actually a good thing. However, when you do actually have an adversary on the world stage, you you cannot continually empower them. So part part of my, you know, ethos about this is that we have insanely deep capital markets. Like, we have the talent. Right? And it's just... This is where it comes down to a lot of people criticize this in terms of, like, you know, communist, like, central planning, basically, where you need, you... You're going to need this kind of capacity one way or another for for medicine, for electronics, right, chips, all of this kind of stuff. We scrambled to to cover like EV supply chains and chips really fast because those were you know critical at that time and it really just turns out that it's all it's all critical. If you forget how to make anything it's bad, right? So ultimately my my whole worldview is that like the toolbox is here. We need we need the people to actually like do it while we're still kind of, you know, working with China and other industrialized nations to produce the things that we want to build.