Limitless Podcast - How AI Brings Manufacturing Back to America | Aaron Slodov, Atomic Industries

Episode Date: June 9, 2025

In this episode of the Limitless Podcast, we interview Aaron Slodov, CEO of Atomic Industries, on the urgent need for innovation in manufacturing. Aaron emphasizes the importance of transitioning from digital (bits) to physical (atoms) advancements, highlighting AI’s potential to transform production processes and tackle the skills gap in the workforce. He advocates for local manufacturing, using the "layer cake" analogy to illustrate investment needs at all levels. Concluding with a hopeful perspective, Aaron introduces the New American Industrial Alliance (NAIA), inviting listeners to participate in reshaping the future of manufacturing.------💫 LIMITLESS | SUBSCRIBE & FOLLOWhttps://limitless.bankless.com/https://x.com/LimitlessFT------TIMESTAMPS00:00 Start02:15 Innovation in Digital vs. Physical Worlds10:42 Shift from Bits to Atoms15:30 Importance of Manufacturing for National Strength20:00 Layer Cake Analogy30:45 Supply Chain Dependencies37:12 Skills Gap and AI in Manufacturing42:55 Precision in Manufacturing51:30 Vision for a Modern Workforce01:02:10 New American Industrial Alliance (NAIA)01:10:00 Call to Action------RESOURCESAaron Slodov: https://x.com/aphysicistAtomic Industries: https://www.atomic.industries/NAIA: https://newindustrials.org/David: https://x.com/trustlessstateJosh: https://x.com/Josh_Kale------Not financial or tax advice. See our investment disclosures here:https://www.bankless.com/disclosures⁠

If you ask the average person about the rate of innovation over the last 30 years, they'd probably say, yes, it's super impressive. We went from computers that were the size of a building to supercomputers in my pocket. And while that's true, that only covers half of the question. That's the world of bits, the digital cyberspace of ones and zeros. The world of atoms is actually a totally different story. If you look out your window, if you look in your living room, not a whole lot has changed over the last 30 years. And I think that's what I want to ask you, Aaron. Is the world of atoms actually making a comeback?

Similar to computing, I do believe that, You know, we've obviously made some great strides over the last few decades, but I do think that this idea of stagnation in the world of Adams, right, attributed to Peter Thiel does have a lot of legitimacy, right? We've, we've kind of neglected a lot of things on this front. And whether it's through regulation, through offshoring, right, through different labor tendencies and trends, there are a lot of very interesting forward momentum carrying kind of concept that we see right now, and I believe that, like, the next few decades is going to shift from

manipulating information and, you know, bits to manipulating matter, basically. So bits obviously gave us cloud apps. Adams gave us semiconductors, EVs, vaccines, right? Like, all kinds of really civilization critical things. But one is margin rich and obviously allows you to grow and recycle capital very fast. So I think that, you know, AI is going to help be the bridge that allows us to kind of bring software scale productivity and like cycles to hardware scale problems. Awesome. Okay. So for the people who are just tuning in, we're talking to Aaron Slodoff, who is the CEO and co-founder of Atomic Industries. And I think a lot of people, Aaron, would refer to you as a physicist, but I kind of want to refer to you as a teacher because what you're doing is you're

kind of teaching machines to build the machines. And it's this whole new way of revolutionizing manufacturing. So what I first wanted to find for everyone who's listening is, is why is manufacturing important in the first place? Why is it important that we get this right? And why is it so important to you to build this yourself? If this is just my handle on Twitter, but and yeah, I did an undergrad in it, but I really believe that, you know, manufacturing is, it's so important because it actually converts ideas into power. right like this is how a country converts its ideas into power and if you can't build you can't defend and you can't lead and eventually you just can't survive because you're dependent on somebody else

right so investing in manufacturing and really pushing the boundaries is kind of when you look at like on a historical civilizational scale this is what civilizations are built on right like being able to to build and produce things specifically and I don't think that investing in manufacturing is not just like subsidization, even though that's how some of our rivals have, you know, gotten to the place that they are in right now, but it's about capability building, like training, equipment, software standards, there's locating the full supply chain for various things within your countries. So when you say investing in manufacturing, what does that actually look like? Are you investing in, in like the factory space? Are you investing in the human capital?

Are you investing in the resources that are required, the raw materials that are required to manufacture thing? What does it actually look like to invest in manufacturing as a category? The way that I kind of view this is like a layer cake, right? And so on the top, you have all the really flashy stuff. You've got like the application layer, right? So drones, submarines, tanks, whatever you want, drugs. That's all the top. That's what gets all the press and the news.

People love the toys. But below that to even make that stuff, right? Like you have a layer of execution, which is. is the people and the know-how and the labor, I think about that as kind of like this precious pool of its own type of resource, right? So, like, if you don't know how to do this, somebody else has to do it.

And then below that, at the fundamental, like, low level is the capacity. So these are actually, like, the machines and the, you know, the actual factories and space and raw material, everything kind of, like, goes in as an input to that process. So in my brain, at least, I see the industrial base as, like, a layer cake

that's kind of divided that way. And investing in it, right, like very similar to computing, when you have this really high throughput of being able to, you know, execute a computation somehow, right? Like, we measure things in flops. We have nothing like that for the industrial base. It's like how many atoms per unit time can you process, right? Like something like that.

That's why I think the idea of the next few decades or beyond just being conceptually about how we manipulate that matter. and what the throughput looks like, really. Downstream of Trump's whole, like, Liberation Day tariffs, I think everyone's just gotten a little bit more educated about the nature of manufacturing, especially also after the whole, like, supply chain issues during COVID, which kind of just revealed how globally interconnected we all were

and globally dependent we all were and some of the costs associated with that. So, like, it's pretty interesting to see that, like, you know, the average bankless listener and the average person in the modern world who follows news is now just a little bit more informed about the nature of all of the, these things. So that's kind of the base of understanding that I think our listeners comes with. And also due to like bankless's emphasis in the world of like macro, we also understand like the Triffin dilemma. If you have the global reserve currency, you have to export dollars and you

need to buy things which naturally exports your manufacturing base. And so this is kind of like all the education that I think the globe has really gotten informed on in late years, which brings up like two main conversations, which is, okay, we now want to reinvest in manufacturing, domestically. And I think maybe all countries want to do this, United States specifically as well, because we once had this. Now we no longer have this. We want it back. But also, getting it back looks different than just investing in it in the first place in the 1970s, right? We're not just building out car manufacturing facilities. Whatever manufacturing is in 2025, 2030, 2040 is going to be different than building out manufacturing as it was. So maybe you can talk about that difference. So like, now, that we are trying to restore manufacturing domestically. What is different doing it today than is doing it like 30, 40, 50 years ago? I'll kind of start from the macro side, right? Because I think a lot of people push back on this stuff because they believe we don't have that comparative advantage anymore. And somebody else has built it. So why should we bother to rebuild our own comparative

advantage again? It's like, yeah, we don't want to do T-shirts and toasters here. And somebody else somewhere in the world is going to do it for cheaper, basically. and people applaud that for some reason and like they've never been to a factory overseas, I guess, right? You can look at this any way you want, but making people slave away in a factory for like 18 hours a day for less than $3 a day is kind of insane. Like nobody should have to really do something like that. Did we just become so wealthy and so, you know, intelligent that we decided to stop innovating and building a comparative advantage, right, on like anything that we've been. wanted to. And being a service-based economy, I think a lot of people believe in that orthodoxy that a civilization is going to trade off these things, you know, like over a long time scale.

But something that I don't think people consider is that maybe that is just completely incorrect or that the economists that actually built those models, they were never necessarily against the idea of being able to reclaim comparative advantage in something, like production. or, you know, like, if you look at the, you know, the first and second and third order tertiary pieces of, like, an economy, agriculture and mining and inputs, right? Like, those can all still be a thing. But to answer your actual question, right, like, I think that, you know, the distance from a supply chain, which I don't think anybody disagrees with this, but, like, distance kills iteration, right? Like, when your supply chain is halfway across the planet, you slow down. lose quality, you bleed, you know, IP. And domestic manufacturing isn't necessarily like

nationalism. It's operational efficiency. It's just like being vertically integrated, right? Like, these are these are very straightforward concepts. And the question, you know, is like, do we want old-timey, dark, dirty, dangerous manufacturing? Like, obviously not, right? Like, all of this stuff has to be rethought. It has to be rebuilt, reinvented, and reimagined in a more like sci-fi kind of way. And when you, you know, when you co-locate design, production, and like all these things, you compress lead times. There's just huge compounding, you know, advantages to doing that, obviously. Yeah, is that co-locating, that co-location, the domestic supply chains, owning your own supply chains?

Is that kind of the predominant feature of modern manufacturing? Are there other features to also bring into the conversation, too? Or is that, would you say that's the main one? I would say that that's a very heavy piece of it, yeah, because being able to figure out how to do something viably, no matter how small it may seem in terms of the overall global supply chain, right, like the T-shirts and Toasters kind of idea, you still need those feeder industries to do anything, right? So when we talk about like drones, for example, right, being able to make the motors and the propellers and the airframes and like the RF chips, right, and the cameras and everything that go

inside of these things. Having the feeder industries to actually manufacture those domestically is important. And you can obviously, you know, big shocker use that capacity and talent for manufacturing other stuff. But right now we're kind of like hamstrung, right? Because we don't have that here. So when it looks like, you know, rebuilding our manufacturing base, do we have to start with those feeder industries? You have to start with the small basic widgets in order to build up to the more advanced widgets. Can we short art cut our way straight to the top? How lengthy, how much work is there left to do to build out our manufacturing base? A lot.

I would say, and there's people on both sides of this, right? Like, America is still the second place, industrial base, right? And first in a lot of things, but also lags way behind second in a number of things. You can go look at, you know, the NIST manufacturing report that comes out almost annually and skim through that and look at it. But I guess you have to kind of go back and look at why we may or may not have outsourced something, right? And trying to understand how to actually bring something like that back

with the amount of, right, like, input material and energy and, like, space and skill required for doing that kind of stuff, right? And the machines. So I just go back to the layer cake analogy, basically. So if you can figure out a way to short, circuit, something like that, and do it more efficiently, domestically, that's kind of the big question, right? So when you look at, you know, Elon producing Starling terminals down in Texas or just rockets or cars, right? Like, there's, there is a way to do this stuff. Do you need that kind of

scale? It doesn't hurt, that's for sure. But, you know, rebuilding these things from the ground up is a different game. And, you know, like, that's kind of like what my company does, but other people are figuring this out as time is moving on. Yeah. Yeah, maybe you could actually just take us into the world of atomic industries. What would you say is your strategy for doing what you want to do? Maybe you could actually kind of give us, like, the general pitch of, like,

what you're trying to do and then also the strategy of how you're trying to get there. Yeah. So, I mean, we're trying to build a 21st century mass production company, right? And if you guys are familiar with this, the underpinning of industrial society is that mass production helps us, you know, build a lot of these things. And to mass produce something, you generally need to, you know, design your product. And then you take snapshots of each different component inside of there. And then you make a mold or some kind of manufacturing tool that pumps out, you know, each of those widgets in a factory somewhere. and the ability to make a mold or a manufacturing tool for a given widget is this like very precious,

tiny evaporating domain of trade knowledge, right, that like that we don't have anymore and we've outsourced forever. So the whole concept of what we're doing is trying to teach AI, right? Like, look at this widget, now build the mold for it and move us from concept to production faster than anybody's ever been able to do. And like we are vertically integrated and the idea of being able to design and build tooling or molds for anything and then put it into production

and start pumping stuff out fast and making sure that these designs are, you know, because they're computational in nature, they're kind of like physically optimal, right? So they're going to run super efficiently. They're going to be designed near perfectly. And the whole process that, you know, goes from start to finish is all done in a factory

that we control, basically. So it's a super interesting way to basically crank the flywheel on the entire physical economy. And that's one of the things that we believe, right, like is a way to short circuit that problem and do it here or do it anywhere, honestly. Could you give us just like the one-on-one of the whole manufacturing process? Because I don't think I've ever really heard that before. So you talked about molds. You talked about building the molds. So I think the molds are the thing that makes the thing, but you guys are also building the things that makes the thing that makes the thing. Can you just kind of walk us through so like you have input materials, raw materials go into a factory, you're building out some of that

middle processes and on the outside, the output is the products. Can you just kind of walk us through the step by step how raw materials turns into widgets on the other side and the parts that you are kind of like engineering in the middle? We basically like end up selling parts to people, right? Like our customers want parts. The input materials are basically like, steel and plastic resin. So they're like little, you know, pellets.

And we take giant steel blocks and cut them up into molds, basically. Right? So like, AirPod case, this came out of a mold somewhere. And the design on this is very crazy. But if you look around your desk, right,

like most of the stuff probably came out of a mold somewhere. And so ultimately, yeah, we have to orchestrate a bunch of machines in a factory to basically go from like the design, you know, chop up the steel, make the mold, put it together, and then ultimately put that mold in another machine, and it's like cycling, right? So every time you open and close the mold, the parts are falling out of it and being transported into a supply chain somewhere. So this is,

this is like the pain of production, right, because that barrier of saying like, all right, here's my product, I love it, this design is perfect, snapshot, now I want to go produce like a million of those a year, right? That whole process takes months and months to actually go from start to finish. And we're trying to reduce that, you know, as dramatically as possible. And the orchestration that happens in between, this may not be a shocker, but it's like, it's kind of similar to a self-driving car, right? Like you take the car as a system and you load it up with software and sensors or you give it vision

and you're trying to teach it about driving around in the world. And what it's really doing is just controlling all the aspects of the car, right? It's very similar. It's like a self-driving factory almost, and you just feed it designs and then out the other side are like being pumped out the parts, basically, that people want. So we're basically trying to, like, obfuscate away, you know, all the horrible pain and suffering in between, you know, design and production. And so what you're talking about is, I'm assuming, what China has gotten very good at over the last 30 years.

They have figured out how to build out this process efficiently without error, I'm just guessing. What are the skills that make somebody good or bad at this? Like, what are the pain points that you are trying to optimize? Is it something like when you're making the mold, maybe you're imprecise and your tolerance. A product comes out the wrong shape. Now you have to start from scratch and build a whole new mold again. Maybe that's a problem. Let's try some of the like the, what are the skills that go into actually doing this well?

Well, this is great because this is like why it's a trade, right? So the humans that do this stuff and that are the best in the world that this to Tim Cook's like point there's a famous video of him talking about a football stadium full of tooling engineers right like in China and and how like America couldn't fill you know maybe like the front row of one section or something like that it's crazy so basically what you have is people putting in the reps to solve and figure out these problems and it's super it's super interesting right because every every new shape that you want to ultimately like have a manufacturing tool for is a new problem, right? So when you think about like language models and if you guys keep up on

this stuff, right, like problems that lie with, you know, they're outside the distribution of the training data are generally like pretty hard for these models to do efficiently, right? So imagine that same kind of concept but for a trade like mold making where, you know, you're getting customers every day, every week with new widgets that they want to build a mold. for. So the more reps that you have, right, in going from that like end-to-end process, from design and concept to final production and like parts in hand, the better. And this is why we, like, in the real world, what we see is people specializing in certain, you know, like industry verticals. So you might have a shop that's very good at like medical devices or automotive

components or aerospace parts. And it's because they've seen the same kind of of like domain or category of part, you know, over and over again. So they're familiar, at least, with, like, how those molds ultimately kind of like function, how the materials that are used in that industry, like work. It's a lot, right? Like, you're juggling a ton of variables. So the people that are the best at this stuff, they've just put the reps in, right? Like, they've seen hundreds or thousands of these problems over decades. And you go to them because they are super efficient at that problem. And, you know, our hypothesis is that capturing a very similar level of

explicit, you know, knowledge around these problems is something that we can use as a general model to be able to do this stuff. And then, you know, short circuit, anything that gets like thrown at us, basically. Yeah, I'm curious to understand how you guys specifically are putting in the reps.

I love the visual of the stadium full of engineers who are very good versus our front row. I'm curious, what is putting in the reps look like for you? Because I understand AI is at the frontier of a lot of what you guys do. and AI has been a tool for leverage. So is it possible for these new leveraged engineers who are just sitting in the front row

to eventually acquire enough bandwidth as the stadium full of these people? Yeah, and that's very spot on. So basically, what's really interesting about this fundamentally at a low level is ultimately, like, you're presented with a shape, right? And you have to figure out how the mold for this thing works. And basically, it's kind of mechanics in a lot of way, right? So it's a very physics-driven problem, and it's kind of a deterministic problem.

But you want to think about it like the solution space for finding the best mold for that shape. This is what these human trades people do, right? Like they use all of their tribal knowledge with a handful of tools, right? Like CAD software and some simulation tool, to kind of zero in on a good enough design. And the problem with this in the real world is that everybody's on a timeline. So these people don't have forever to sit there and iterate and iterate and iterate on designs. So if you take a physics-based approach to solving this problem, right, like you can shrink down that solution space pretty rapidly and then iterate through it with software, basically.

Right. So we can use, you know, deep learning and a bunch of other very interesting machine learning techniques in AI, right, to zero in on these more optimal design. right and you're just like hey I want it to be like this price and then it goes and finds that design right and like oh by the way now that it's on a computer you can do thousands or tens of thousands of these things like every day right and like in the real world it's one person at a time until they're done it's not like a google doc where you know a hundred people can get on there and make it go faster so what does that look like you mentioned there are some problems that are kind of outside

of the context space surely you're doing more than just interfacing with chat chbtee and saying hey Like, how do I create this mold? Is there, is there like this reinforcement learning that's happening where, where you kind of teach the AI as it goes what's good and what's bad? Like, how does, how does AI actually play a role in creating these new tooling? Yeah. So this is where it comes down to kind of being like a full stack, vertically integrated problem like self-driving cars, right?

So you have to be able to teach it when I'm designing, like, you know, a water channel inside of the mold to cool off the mold. where like how do I actually drill that hole or 3D print that hole so you have to take into consideration the manufacturability of like a feature or something and it's like once you once you build up enough knowledge around if I design it this way I violate the physics of that problem or like the drill breaks or something right so you do you kind of have to you have to start at like a first principles approach where you're taking heuristics and tribal knowledge from people that have done this for a long time,

plus, right, like the physics of the actual problem. And it's not, it's not today, at least, like, the domain of a language model. So we don't technically use, you know, these foundational models from the big AI labs to do this kind of stuff. We have to build

it from scratch and kind of, like, from the ground up, teach it, like, where things are violated, basically. So if you're, you know, you're going through the process and like dropping it, dropping and this thing is like a CAD file, you have to be able to, you know,

understand no, like the drill breaks here, or like the metal won't move this way or go this way or it gets too hot. So there's a lot, there's a lot, right? Like, there's a lot of different weird problems that, like,

very analogous to how humans learn it. We're teaching the software at the same kind of rate, but it's a, it's a single training loop, right? So we kind of like teach, teach our software one time and that's going to be good enough for the next, you know, 500 years kind of thing. And it will only improve the more of these jobs that we do ultimately

because we're collecting data from the factory floor, right? Like the CNC machines and the drills and the 3D printers that we use, we take all that data and feed it back into the design software. And then ultimately, you know, when the mold is running in a factory and spitting out parts, we can collect that data too, right? Like, what are the times look like? What are the temperatures and all this stuff? Like, all those things can be collected and fed. right back in. So that's our version of like the self-driving car model. It's funny. As I'm hearing you describing this, I'm drawing ties to this video that I saw last week with Tesla's Optimus Robot. And it was this crazy robot that was running around dancing, but it was the first time it was ever

actually dancing because it had never been taught this in the real world. It was all trained on this synthetic data in this virtual reality world that they created to train it. And I imagine there's a ton of cost efficiency there. So when I'm hearing you talking about putting in reps and getting as many iterations as possible. Is that something that you guys also do? Are you able to kind of get these reps in in cyberspace versus the real physical world to save a lot of money and increase that leverage? You kind of mentioned something similar to this a little bit ago, but like you can, but there's a catch. The thing that matters the most specifically for this, because we're talking about production engineering, right, where the tolerances actually matter. Right. So you, you, like,

The one thing about language models and like RL for robots and stuff like that is that currently there's room for error, right? Like if the robot just moves like an extra millimeter in like one direction, like that's okay. You're not going to like the whole factory doesn't go down. But when you're when you're actually like designing and fabricating one of these, you know, manufacturing tools or molds, yeah, it's like sub five micron tolerance levels. right? So everything that gets spit out of the software, there's absolutely no room for error at all. How big is a micron? It's a millionth of a meter. So, you know, I don't know, in wits of hair, I guess it's probably like, what, half of a half of a hair? Yeah, I think so. Very small. It's very small. I could be wrong about that. I think I'm going to get so grilled for being a physicist in saying this stuff. Yeah, cut that out.

Yeah, I think it's, I mean, it's tiny, right? And like what the problem that you're dealing with, especially for something like this, like you guys probably have AirPods cases, but when you look at it right here. Yeah, the design and just like how crazy the angles on these things are, which Apple is famous for. Imagine this coming out of, you know, like two pieces of metal. Like that has to mate together perfectly. And there can't, there can't be like a seam there. And that like you guys have probably seen this on a lot of plastic stuff that has

ever been made. You see like that little seam or like the weird little like dot. Like that's where the plastic kind of, you know, entered into that. It's, it's crazy because you have to like hide it. It's a, it's aesthetic. And it's a really interesting engineering problem. So yeah, I would say you can't really like compromise on stuff like this. So having it be a physics driven approach is, you know, for the time being at least like how we are able to solve these problems. and, you know, tying it into what actually happens in the real world and building this, like, product, you're building predictive capability, you know, over time.

It's the same thing as, like, driving the millions of miles in the car. Yeah. So we have this visual of a whole entire football field of Chinese, like, manufacturing engineers. Are those, is that the right title of manufacturing engineers? Tooling engineers, yeah. That was taught under, their ships were sharpened with, the iPhone, which is a very precise, very good thing I would guess to get, like, trained on.

And now this is kind of what we are competing with here domestically inside the United States. But the cool new trick that we have is we have AI. And so instead of having to, humans learn, at least I learn over multiple iterations, like I need to learn the same thing like four times before it becomes knowledge in my brain. And that's just experience. That's just kind of how it goes, trial and error. AI has a much faster cycle. Like it doesn't have to fail four times.

It can collect data and adapt its behavior much more quickly. And so I guess that's one of the big advantages that we have in restoring manufacturing in 2025 and beyond. And so that and then also just like, I would have guessed, the toolings are just more advanced these days. But that's really the huge competitive advantage is we can just learn faster. Maybe we don't have the people, but we can just learn faster using AI in order to produce more dependable, more solid, more precise outputs. Where are we on that arc of training AI? Is AI a burden for you because you are still training it? Is it actually starting to become helpful for you now? Like, where are we in that journey? It is kind of a burden because people can still do many of these

things like order of magnitude a little bit faster. So we want to augment, you know, people, basically. And the idea for the next like couple of years at least is can, can the A.m. that we're building make, you know, like one American mold designer, right? Like, as effective as 100 Chinese mold designers, right? Like, or productive as one. So we, we have this very interesting in-between moment where it can do a lot of the unnecessary kind of repetitive, you know, work that designers don't really want to do. And, like, even internally at our company, they're like, can you guys just, like, finish

this magical AI, right? So we can design more. We're like, yes, we're working on that. So it is, it's super interesting, right? Because, like, I think as soon as you play with one of these things, like, if it was a chat GPT or, you know, like what we do, there is that kind of like that chat GPT moment where you're like, this is crazy, you know?

Like, I can be orders of magnitude more productive using something like this. Ultimately, the faster we can build and train this thing, the better. and I think that that's not just for like what we do but for a lot of things right like taking this very precious tacit knowledge that people have and being able to preserve it replicate it scale it and then like you know building a new workforce on top of that I think is it's unbelievable and it's super powerful and I think that's kind of another way that we can heavily incentivize right like people going back to work in this stuff because now you're kind of like an orchestrator of AI and And there's people that'll push back and be like, yeah, well, you're destroying, you know, all the underlying skill and the knowledge that people used to have to put the reps in to build.

And I would also push back on that too and just say, right, like, there's nothing stopping them from also learning the fundamentals while they're, you know, learning and being augmented through an AI like that. There's no reason you can't do both, right? Just because I use chat GPT to write an essay doesn't mean I can't also learn how to write. So I think that a lot of these things can go hand in hand to train a workforce, you know, way faster. So instead of 20 years, how about just like two years, right? Like teach people the fundamentals, like let them be augmented by these things and grow and scale really fast. And now that you're kind of like a superhuman able to do this stuff, you get paid more, right? It's amazing.

There's tons of benefits across the board. And you get to work in like a sci-fi factory too. So, yeah, I think it's a really powerful thing because people love actually making things, right? Like, there's a very, I don't know, natural, implicit kind of idea of building something with your hands and seeing it, like, pop out the other side. So I guess that layers on top of the previous benefit that we talked about, which is the supply chain. So if you can eliminate the global supply chain and all you have to do is transport materials domestically, that's already a huge cost savings. then the second layered on top of that cost savings would be the classic example of just like, well, it's not going to be somebody, AI is not going to take your job.

It's going to be some high performer who's using AI who's going to take your job. And then you layer those two things on and then we have a domestic manufacturing revolution here. I mean, like, this is the thing that we talk about at bank lists with our writers. It's like using AI as an expectation. You are expected to use AI in order to be the best performing person at your job. And this is like, and it become like a 100 to one like ratio in terms of leverage on your on your work. And this all got started, of course, when Gary Kasparov lost it to Deep Blue in like the early 90s or something. And then what happened next was the birth of the AI plus Grandmaster versus AI plus Grandmaster chess world.

It's always AI with a human rather than just AI. And so we're taking that. We're applying it to manufacturing. And so like, I would guess like, what kind of products are. are you making today? Like, where is this AI process engineer, this manufacturing engineer, plus the human? What are the materials that you are building today that are being manufactured onshore? What are the levels of sophistication that we're doing right now? Yeah. A lot of the stuff specifically, I can't really get into too much detail about, but if you think about different

industry verticals, right, like automotive parts and components, stuff for consumer across consumer electronics, I mean, CPG goods, right? Like, everyday kind of household items, there's a ton. Industrial components, medical devices, and then, you know, defense, obviously. And a lot of these things are just standard components that go into products that you use every day, right? You just don't know where these things actually come from ultimately. So what we're seeing right now is even before the electrical, right and to your very early points right COVID set a lot of the stuff off I think when supply chain

became a household term right when people were running out of toilet paper and like basic stuff it freaked a lot of people out and it really did put a huge spotlight on the sensitivities of the global supply chain right and do we do we actually need to eliminate the global supply chain and just move it all domestically like no not necessarily but I think that having an approach like this allows us to re-select, right, like a level of globalization that we feel more comfortable with moving forward. Because I think a lot of people did not expect this type of outcome that we are finding ourselves in, right? Like, you see these, like, massive sensitivities in the supply chain, but then at the same time, we have enriched our ideological rival, right? Like, over the last four decades. And now they are, you know, knock,

at her door or at least saying that they're going to just do whatever they want. And that's kind of interesting and frightening to a lot of people. But yeah, I do think that this kind of stuff, moving forward at least, is going to be super powerful for a country to like reclaim, you know, more of its sovereignty and its ability to also produce what it needs and wants and be a little bit more independent. And yeah, I mean, the opportunity is ridiculous. I don't know. It's huge. Manufacturing as a total of GDP is like, yeah, $14, 15 trillion, right? Technically larger than, I think, tech as a whole. But yeah, these markets are gargantuan. And if you do something like this where you're compressing, you know, the dev cycles basically from, you know, months and years to

days and weeks, you basically create a new economy. It's very similar to what AI is going to be doing in the digital world. Same thing for the physical. I definitely want to talk about the TAM of that and really how big all of this is. Before we get there, though, I do kind of want to understand like kind of where we are in being able to like access that TAM. So I think we're still, we're still like early in this whole, you know, combination of AI in manufacturing. We're doing some of the high value stuff, the national security stuff, the things that are non-negotiably must be. be done inside the United States. The high value stuff that's important. And then I think as you build out the scales and increase the capacity will start to go down the tail and into the long tail

as this thing kind of scales out. And I think I see like two different vectors in how that happens. You get better at what you do. And then you also do more of that. So like qualitatively and quantitatively, where are we on those two kind of like spectrums here? Like how how quality, like if we're working up the stack of qualitative measures, how far up are we? Are we? and then where are you in trying to like multiply your work horizontally? Yeah, I mean, we're just in the beginning of this stuff right now, right? And this is something that like we have to hoe the line on very carefully when we're actually building one of these companies or one of the first of these companies.

It is very, very early days. So you have to do exactly what you're talking about, right? Like you have to show people with the high, high value stuff. And then you can kind of work your way down. the value chain into these much, much larger opportunities, but greater a number. But the idea of being able to use your capability as a way to, you know, extract more margin out of something that typically in an inefficient market, right, like, is hard to extract margin from. And so, like, a lot of the processes and, you know, shops and people that are doing

stuff in America today are still working more on those like, you know, those easier targets or more complex things that they can extract more margin out of. And I think, you know, throwing a layer of tech on top of that allows us to kind of like scoot down the value chain a little bit more and open up that tam, like what you're what you're talking about, basically. So I think you'll kind of see more and more people taking these like vertically integrated, you know, whether it's with AI or just like being software defined, you know, from the ground up approach to, you know, the thousands of different industrial processes that make up the industrial base and like in that $15 trillion market, I think you'll see it like start to

saturate more and more over time. And the idea of being able to, you know, demonstrate it on like a small scale and then go to people and, you know, try to get their help to scale it, I think the optionality will open up there on how that unfolds and who, you know, is ultimately backing and helping to build those companies. And yeah, doing that right now is, it's tricky, right? Because like, we have to use venture for that because you can't just walk into the bank and be like, yeah, give me, give me a $10 million factory and let me try out this like AI thing. So it's tricky for sure. I'd love for you to walk us through the actual specifics of this addressful market. Because I think a lot of people listening probably don't understand quite how

large this thing can get. An example I love is Apple when they released the iPhone, I think, was like $70 billion. At the time, I think it was Exxon, that was maybe at $350 billion. And if you were looking at Apple at that time, that was probably the perceived cap, it's about a 5x, because surely it's not more valuable than oil. That was the most valuable thing in the world. Fast forward, they reached $4 trillion in market cap. So I'm curious the downstream effects in the size of the market, as we get another one of these large unlocks, as we get humanoid robots at scale that can replace the productive workforce of the United States or as we get drones and delivery vehicles and just a lot of new ways of moving information and moving value around, how large does that market get?

Are we facing a similar thing where maybe the ceiling was three to four trillion and now this next ceiling becomes even higher? How big can companies and can GDPs grow from the downstream effects of this? One thing to look at, right? Look at the market cap of Foxcon, right? As an example here. and I think, I don't know, what is it? 75 billion, maybe 100 billion. And they almost have like a million employees now. Or BYD even, right? Like look at BYD,

Xiaomi, any of these companies. So Foxcon is an interesting example because I think more than 50% of their revenues come from Apple. So when you are basically the backbone of Apple, what does that ratio look like, I guess? So I do think that that's a, it's a fascinating concept, but even Foxcon, right? Foxcon can only produce as much as, you know, like infrastructure they have in human and machine capital.

Right. So when you, when you see these announcements of like BYD trying to build a 50 square mile city that's basically just like a huge factory city, that's crazy. Right. So you need, you need some kind of lever there. where the human capital is getting traded off for, you know, software or automation. But in general, I think China looks at the ability to use its human capital as a huge lever, whereas, like, we can use it, you know, with an augmentation of AI to be an even bigger lever. Obviously, we don't have, you know, a billion people in America.

But I do think that ultimately, yeah, the market caps and the, the opportunities for these companies are kind of like a function of their machines and their people. So like if Foxcon could produce everything for Apple, Samsung, and like, I don't know, what's another gigantic brand? Like, how big would it be?

And how many people and machines would they need? And could you offset that somehow by like, you know, an AI that knew how to do production? And could you make those cycles faster? and I don't know, would you just be like spewing and pumping out like consumer goods to the world? Like everybody can now have an iPhone because it's that cheap, right? Like maybe an iPhone is, I don't know, like 100 bucks in 20 years from now. Like those devices will become commodity or something, right?

It's, I don't know, you can take this stuff to the extreme if you really want to. But I do think that manufacturing as like a concept is kind of a race to the bottom always. right? And that's why we kind of find ourselves in this situation. So it'll be, it'll be a really, really interesting thing to see unfold, you know, like this new wave of industrial companies, because I think that the more they can handle and then produce, you get this weird, you know, Jevin's paradox of like, now we can produce more, people are going to consume more, yada, yada, yada. And like when, when AI is happening on the digital side of this too, right? So, like, does the economy actually get bigger and keep growing at this explosive pace?

Do we actually leave the planet, right? Because now you have to produce stuff for space and, like, other worlds. I don't know. It's crazy. Just like the compounding multiplier effect of all these things. And then you need energy for that, obviously. Energy might be like the next limiting factor ultimately. That's what seems like.

So as we wrap up this section, I'm curious to hear a little bit about the research. since why we haven't quite been accelerating as fast as we like in terms of manufacturing. I have this question where if Congress handed you a one-time red button, it's a, you've raised or rewrite one federal rule, and in doing so you can unlock a trillion dollars of factory output within five years. How would you address this? What would you create? What are the bottlenecks and the thresholds that are kind of in the way that if we're out of the way, could accelerate this even faster? Do I have to only have one bottleneck or do you? No, no, please walk us through however many you have. So, all right, I'm going to go back to the layer cake example, right? I think at the very

fundamental low level, we need, obviously, like, we need energy. So energy has to be commoditized, or at least we need to have a surplus of energy. And, you know, it would be great to have economic zones with, you know, production capacity, and they're all just, you know, directly, like, tied to a really heavy grid where we can draw on that power, whether it's like having a, you know, a nuclear reactor, like in your factory, just powering that factory, like a data center, right? Like the same thing, but just for mass scale production of whatever. And then obviously, like, raw materials.

So we need, you know, we need an explosion of mining again. And I do think that that's super important. But like, yeah, energy, raw material is huge. And then the other two pieces of that layer came. So on the capacity side, we don't really have the ability to make machine tools here. So like CNC machines and lathes, right? And like the actual tools and machines that we need, those basically all come from somewhere else today. So I would love to see somebody retaking, you know, that problem and blasting it into high gear.

Like I'd love to see basically like Apple applied to industrial machinery. I'm like all software-driven, super modern, flexible machines, and I'd love to see an operation warp speed for machine tools, basically. And then on the human capital, like, you know, upper level, on the upper level there, I'd like to see being able to incentivize people to go back into this stuff, right? So rethinking labor laws, right? Like the way it is now federally, I mean, if you touch a machine, you basically have to be like an hourly employee somewhere, basically.

There's a lot of really arcane rules that were meant to protect people, you know, during the last industrial revolution that we need to like rethink. So I think that there's, yeah, regulatory stuff that we can work on in terms of like labor laws. Also, plenty of other stuff on the regulatory side, too. But I do think that, you know, export. controls are something that's really interesting, certifications, you know, requirements and quality, like stuff like ITAR, right? That kind of stuff. And then stuff that's not necessarily

like subsidy, but just rethinking like tax, tax laws around standing up capacity and capability. So incentives, not just for like buying machines, but maybe, I don't know, like, remember opportunity zones? Maybe like opportunity zones, but for manufacturing. manufacturing very specific ones right like stuff like this so there's a lot of i think there's a lot of really smart things that we can do that don't necessarily mean that like dupont is going to be poisoning your backyard again right like for the next 50 years but allow us to like move faster as well so yeah there's a there's a lot there but that's kind of how how i would wave my magic wand yeah yeah there is a lot there and it comes from all over there's some regulations some bureaucracy stuff

that you talked about, there's just like the whole need to bootstrap our own companies in manufacturing base that bootstraps the rest of manufacturing. So there's like second order bootstrapping that we have to tackle. It seems like a lot. It seems like it's very hard. Are you optimistic about the United States ability to do all of this? Because it seems to be that there's so many things to do. And a lot of the, they're all hard. They're all hard. It seems like every single frontier that we need to go in. in order to do this is a hard frontier to push. So what gets you motivated? Are you excited to do this? Do you dread it?

Like how optimistic are you that we can even do this? Like, share us a little bit of your sentiments about this. I don't think there's an option, first of all, right? And secondly, I think you need people like me and other people like us that, you know, like Elon has a quote from somebody, I think, famously about building companies, right? It's like chewing glass. you want people that like to chew glass to do this stuff because to your point, right, it's hard.

It's not fast. It's not easy. And I am super optimistic about it, right? Like, nobody has come to the table with this much force about, you know, bringing American manufacturing back as a whole for a while. And I do think it's like reindustrialization is one of the most interesting, like, unifying fronts politically that, that we all have. And yeah, sure, people have different ways of thinking about doing it.

But I don't think that people necessarily hands down disagree with this concept. So I do think that, like, I derive a lot of optimism from, you know, talking to people every day about this stuff, right? And, like, you know, building this company, seeing the support that we have seen out in the public sphere, you know, a lot of what we do in, in this realm at least so far has been super exciting and seeing all the companies that are popping up and trying to, you know, add momentum. I think like the more of these vectors that we like throw down and add, it's just like it's not going to stop. And I love that. So I think, yeah, I'm super thrilled about the future of what this looks like. And I think, you know, building one of these companies firsthand too gives me a lot of promise because I see it every day, you know, when I first started

I didn't know that this was going to be possible. I mean, like, I was going to make it possible somehow, but yeah, making it happen every day is like super, super exciting. And I don't know. Yeah, we can get into more of the stuff that we're doing on the backend. That's not part of the company. But yeah, I'm stoked. Yeah, I think we, I know what you're talking about. And I also do, I think we want to go there next.

But I do want to kind of paint a picture for what this would look like when this all kind of comes. into formation and it looks like a reality. Like, what lifestyle would the average American be living in 20, like 20 years? Everything on Amazon is like cheap or free. We can build things after drawing a dumb in Microsoft Paint after two seconds. Like, what does the lifestyle look like once this like domestic, AI-automated, manufacturing base becomes extremely sophisticated in very high capacity? Yeah.

I mean, I do think it is, I mean, it's, I don't know if it'd be like paint to, you know, your hand in two seconds, but it's like, low fidelity. Yeah, I do think that the entire pace of our world, you know, dramatically increases. And I think that our ability to

you know, to have more comfortable lives and like this idea of abundance, right, like in that new book, but even before that, right? Like a lot of what Sam Alton says, like having an abundant future relies on something like this. Right? Yes, like the language

which models are great, and they kind of act as this digital brain that with enough data, you can kind of teach it to do any kind of digital task, right? But most of what we ultimately have is an industrial problem. So being able to live in this new world of abundance is kind of what you're talking about, right? Like the cycle of going from A to B on any industrial process is like dramatically shortened, we don't have to worry about these things anymore, or at least like worry about them as much. And it really opens up, right, like a lot of new possibility. Because I think we can actually rebuild, you know, the middle class with a lot of this stuff. And I don't think people realize it because the focus is so much on like the factory job, right?

Like maybe a factory doesn't need a thousand people in it anymore. Maybe it needs like a hundred. But, right, like the up and downstream effect of that factory. factory job has as much or more impact on the economy as it used to, right? Like, I think a pretty standard, well-understood multiplier effect of manufacturing in the economy was that, like, A, it has the highest ROI of like, you know, dollars invested back to the economy, like almost three to one, which is crazy. And then in terms of like productivity and labor multiplier, one factory job created seven other jobs, like up and downstream, right? Like, whether it was logistic, the retail, right, like quality inspection, whatever,

it's very interesting, right? So like do sci-fi manufacturing jobs, multiply that even more, right? Because if we bring back raw material processing, energy, like all other stuff, and we start exporting goods too potentially, right? There's a lot. There's a lot of work to do. And I think, yeah, it is going to be hard, but I think that looking forward that the country in the middle class

and a lot of America could, you know, could really stand on this as kind of a new, a new accessible industry for a lot of people that doesn't have like a super high barrier, right? Like the technology does. And if you guys, I don't know how old you guys are, but like, you know, growing up from the 90s to now and seeing just like how big tech and technology in general

has like transformed the economy, right? Like people being able to study computer science and like go off and get a crazy, paying job out on the coast somewhere or just like go into a non-traditional industry that didn't have technology and applying that there right like in finance or something it's i think that's going to happen again if we are staying on this trajectory yeah we've we've uh i've seen that optimism a lot from you and i really believe it and and one of the things i admired when we were looking for someone to find as this categorical defining episode uh we landed on you because of the the optimism and

drive that seems very genuine and relentless towards pushing American manufacturing forward. So first, thank you for that. And second, it led me down this rabbit hole where I found this thing called the New American Industrial Alliance. And that is something you created. So I'm curious if you could just tell us a little bit more about that. Sure. So last year, we put together a summit in Detroit. And the idea of this thing was answering this same kind question, right, that I've answered probably hundreds of times over the course of building my company, which was what does this future look like, right? Like, people want to know across the military, the government, industry, finance, tech, everywhere. And I figured it would be a good time to take all

of these people at kind of the stakeholder level or above and, you know, smash them all into a room together for a couple days and start building community around this and common understanding of where the future is pointing. So last year, I was lucky enough to have like a crack team of people, you know, join me on this crazy journey. And like in less than three months, we put together a grassroots, you know, event in Detroit, had thousands of people like RSVP to it, had, you know, we brought 800 people in, had amazing people sponsor it. But that's what it turned out to be, right? It was this really awesome big tent thing that we put together on short notice and people loved it. And one of the takeaways, you know, coming out of that was, hey, we want to actually double down on this, you know, momentum, on this moment and build something around it that's, you know, durable and can endure, you know, more than just like a political cycle.

And the idea of what NIA or, yeah, the New American Industrial Alliance was is that, you know, we wanted to establish a coalition of, you know, founders, build. builders, operators, you know, brothers in arm, trying to basically like unfuck the American industrial base. I mean, I don't know if we can lot that out, but we want to put everything forward, right? And we don't want to, we don't want to wait for bureaucrats. I mean, like, even where we have to, but we want to do our best to have a big tent effort and do it from like the bottom up and yeah, have, have, like, everybody on board. So, like, having the startups is kind of the tip of the spear, and then all the kind of, like, the heritage and SMBs

and, you know, the rest of the industrial base along with us. So it's something that we started kind of on the back end of that summit last year. And in, you know, a year's time, we've definitely made some great progress. And looking forward to, you know, kind of unveiling what we've done over the last year because we're doing that summit again in Detroit in July. And it's going to be, it's going to be great, very much looking forward to it. That's super exciting. For the people who are curious about this, who want to kind of get involved, what type of skill set is required for manufacturing? Because you mentioned, we have a shortage of tooling engineers, but perhaps software engineers now can kind of

fill that void by using these tools like AI, what is the skill set that someone like you would be looking for in a person? If they wanted to get involved in manufacturing, what do we need to more of. I kind of like to explain this in two different buckets. One is basically I need to make a factory to make something like now. So the people required to do that

obviously have that skill already and can be kind of like plugged in to be able to do that, right? And the second bucket is I want to reinvent the factory. Right. So yes, we obviously need people that have some of that skill of the old factory, right? And like old

process, but we need to bring in the new as well. So having a fresh set of, you know, eyes and skill from software engineering, because I think these cultures are, they're very interesting. They have a lot of commonality, but they're also very different, right? I think a very interesting common characteristic of either, people from like manufacturing or software engineering or tech are people that just like want to grind, right? And like mission driven. people, especially ones that like to get their hands dirty, I think those people thrive in something like this. And it's honestly crazy because you have people that, you know, might have worked at Google or something and maybe they had like, you know, a CNC machine

in their garage or they had like a whole wood shop or something, right? Because, but having, having people that are super psyched about, you know, actually getting their hands on steel and seeing like, how their engineering capability can transform a process like this is super powerful. And on the other side, right, like people in manufacturing, they work their ass off. So there's no shortage of people that, you know, grind really hard there. It's more just like the people that want to see and embrace technology kind of like, because for a very long time, these worlds have been like oil and water, you know. It's really, really hard to get technology to make a big impact on manufacturing, at least

traditionally, right? Like, we've just been, we've been like layering in software, just like slapping software onto a factory or something. But it's, I think we are, you know, like cresting that wave of showing people what like software defined manufacturing and like AI powered manufacturing is going to look like in the future. I'm curious if you, do you see a shift of trend in the workforce? because as I'm hearing you talk about this, like earlier on, childhood days, I was building little rocket ships, the model rockets, and doing a lot of things in my hands. And then over

time, the technology just kind of got so interesting that it was hard to pull myself away from it, where there was a lot of really cool things happening in the convergence of like software and hardware. Now we have AI. And it's kind of nerd-snaped me in a way that was all-encompassing. But now that I'm hearing this software is being applied to the hardware world in a way that's kind of accessible and exciting, it feels like not listening to you speak. I'm like, man, going to a factory and building really hard thing sounds pretty cool. And I'm curious if you, if you see this shift happening, are there people who have sat behind a desk their whole life who are now really excited to actually get in a factory? Is that something? Because traditionally,

the, like the big tech has this connotation where people aren't really working super hard. They're just kind of getting by and they're shipping code. And how does that shift now? Yeah. I mean, that's 100% right. I mean, it's what when you, when you can start demonstrating to people that, you know, this type of tech has a direct impact on, you know, you walk into a factory and see these huge machines just like chewing through metal. It's, it is kind of crazy because it's stuff where it's like, yeah, I never even knew that this happened. And now, now you can basically go from like, you know, your laptop in another room and then like go walk out on the factory floor and see like your stuff getting spit out the other side. And you're like, wow, okay, this is like, this is a real,

a real thing. And I think that these factories are, the more and more of them that we see, they do kind of turn into these like kind of cathedrals in some way. It's like, it's a, it's a very interesting experience, you know, because you kind of like, you're taken aback by the scale and the size and the awe of it, right? Like, I can't even imagine going into like the Tesla factory, you know, like the gigafactory now in Texas. That thing is, it's absurd. I do think that I've seen a huge, you know, trend in this stuff over the past, like, two years. And the more, this is going to sound silly, but I mean, I do think the more, you know, the more content that's out there, the better.

And I think that, you know, showing people, because this stuff is so visual, right, in, like, in a lot of ways. It's, it's one thing to talk about it, but when you actually see it and, like, get to put your hands on it, it's just, it's a totally different, like, visceral kind of experience. and I do think it captures people. It's really easy to snipe people at this stuff. Well, and to your point, the scale, like, we have Gigafactory and then just recently Starbase is now an official city.

So you can go to Texas and you can visit a city that is built around a rocket-making facility. So the scale is growing really rapidly. And when you hear things like this, like, oh, now there's a whole city just dedicated to building rockets, that's pretty badass. That just seems really cool, something people can get really excited about. Yeah, I think so. I think obviously there's some kind of. of, you know, like, hedonistic treadmill for all people, right?

So if you get exposed to enough of this stuff, you just kind of get burnt out and you're like, yeah, whatever. But I do think that, like, this is one of the really rare exceptions that it just keeps giving, you know, like from little kid all the way to full grown adult that's, I don't know, 80 years old, like seeing a giant machine. It's just like such a cool thing. And I don't think, yeah, the novelty of that never really wears off. And it just, I don't know, I think it helps show us that we are capable of building these, like, crazy things.

And that capability is, like, only going to get stronger and stronger over time. Yeah. Aaron, the timing of this episode, I think, is pretty good because the episode that we did right before you was with Isaiah Taylor talking all about nuclear energy and the prospects of all having a nuclear energy box inside of our homes to plug in. And I think it's pretty easy to visualize the idea that one of your factories or, one of the factories that you help spawn is powered by nuclear energy on one side as the energy input to allow for us to build anything. And then on the other side, the output of that, we haven't done this episode yet, but we're doing an episode with the founder of Zipline, who's trying to just allow drones to take and transport all goods all around America. So you could just totally

imagine a nuclear energy powered mass 3D printing, 3D printer machine that output hops into a little drone and then comes to my doorstep after I purchased buy on Amazon like three hours earlier. So you can kind of see the whole supply chain starting to come into existence here. So thank you for joining us on Limitless Today and help illuminate your part of that domestic supply chain that is helping us bring forward the future. Yeah, for sure. Thank you guys for having me. It's awesome. Yeah. And also where can people find and support either you or the NIA or atomic industries? Where should people who are interested try to find you at? I'm on X. My handle is a physicist, and NIA is that newindustrials.org and atomic is atomic dot industries.

And yeah, we'll be out there. All right, everyone. Thank you for listening to the Limelisk podcast. And now we know a little bit more about the future of domestic manufacturing here onshore. Aaron, thank you so much for joining us here today. Thanks, Dave. Thanks, Josh. Thank you.