What people don't understand is AI is a technology. It's not a product. And even a chatbot is not a product. It's not been productized.
Welcome to Go No Go. My name is John Bruner. This is a brand new podcast where we talk about technology, hardware, manufacturing, design, and the calls that can make or break great products. It's brought to you by Lumafield, where we spend all day looking inside amazing objects from every industry, from athletic equipment to medical devices and aerospace components. And in this podcast, we'll bring you a lot of the stories behind those products, talking to people who have delivered great products and asking them to reflect on how they designed them, how they manufactured them, and the impact that they've had on the world. In today's episode, we're going to start with an interview with Tony Fadell, the father of the iPod, the iPhone, and the Nest Learning Thermostat. After that, we'll talk through some recent news from manufacturing and supply chains. And then we'll close with a segment called Reconstructions, where we revisit a historical event that changed the world of hardware and manufacturing. Our very first guest here on Go No Go is a legend in the product development space. Tony Fadell is the creator of the iPod, the iPhone, and the Nest Learning Thermostat, as well as an expert on product development from a variety of other experiences. Tony, it's an honor to have you on.
Oh, it's wonderful to be here and wonderful to be here again. So, thanks for the invitation. Hopefully, I have something interesting to say again this time.
We're all big fans of your book, Build: An Unorthodox Guide to Making Things Worth Making, and we have copies of this on everyone's desk here. So, one of the first things that jumps out when you read Build is that you write about how vision and detail are both essential, and normally these are kind of at opposite ends. You have, you know, product people who are vision people and product people who are detail people. Looking across your experiences, you know, at General Magic, at Apple, at Nest, at the companies you're advising now, what's the through line? How do you bring those together?
You know, it's something that I've acquired over time, right? I used to be so much in the detail, really the engineering details of things, that sometimes you forget the big picture from an engineering perspective, but you also forget the big picture from a customer perspective. You don't even know what that is. You're just like, I'm going to engineer this thing, it's going to be the best thing ever, and everybody's going to want it. And so, you know, a lot of engineers start that way, and then they find out, wait a second, not everybody thinks like you, and not everybody, you know, you have to really translate it and understand what the full audience may need to hear to be able to get them engaged and moving forward, right? And so what I like to do now is, through those very hard lessons, make sure that you're balancing both at all times, you know, and you're always talking about the why, the why, the why, and then what?
Look at that. Isn't that cool? Isn't that cool? And that's what I love about Lumafield. All of these great products are literally much more than the sum of their parts. I mean, it's all about the integration. And I think one of the breakthroughs in a lot of the products that we're going to talk about today have to do with the integration of software and hardware too. You know, these used to be separate disciplines, and I think that a lot of the products that you've brought to life combine them in a very forward-looking way.
Oh yes, systems are absolutely important, and it's systems beyond the box, so to speak. It's systems that have all the network components into it, and maybe multiple products all connecting either together peer-to-peer or from a central kind of server thing. So you really have to look at all of those pieces. And luckily, with both the Lumafield technology as well as now AI to help you, you can start to really see how things are put together if you have open source to see how things are put together.
Right, right. I want to take a little bit of a turn into manufacturing here. So the iPod and the iPhone were both big design leaps, and they were manufacturing leaps. Can you tell us a little bit about, you know, what you learned about factories and supply chains and the reality of knitting everything together at that scale?
You know, it's so easy to replicate electrons, and it's gotten easier and easier and easier over time, right? But atoms are getting... well, in some ways it's easier than it was before. But because we have such design tolerances and such volumes, and we're really pushing the edge, it's getting harder and harder to scale those things up because every nanometer can matter in some cases, right? And so you're always playing this game of how tightly can I get, how small can I get something? How many can I make of that, and what's the yield of that? What's the price of that? And so you're always juggling all these parameters to try to make that work. And so in the earlier days, if you look at the original iPod, it had so much air in it. Like it's crazy, like you're like, 'Oh my god, we just threw that together.' You would never ship something like that today. It looks like in some ways a Kickstarter project.
Yeah. It has literal wires inside. You don't see those inside Apple products anymore.
Exactly. It was just one board, literally it looked like a Kickstarter project back in 2001 now that you look at it. And then you fast forward, then the iPhone, it had a bunch of air, the original one. And now if you look at the iPhone Air, it's like, give me a break. There's nothing left. Like it's integrated to the... the antennas on all four corners, like the battery is the structure, it's crazy to think what we've done in 25 years to unify it all. And so that's something on the high volume, and now you're making tens of millions of them, hundreds of millions of them, and to get that to work across... remember, you're not just talking about one design.
We're talking about multiple battery vendors, multiple mechanical vendors, multiple chip vendors, multiple display vendors, multiple PCB vendors and flex vendors. And when you think about it, all the different permutations of all these things coming together with all their own design tolerances from all these different manufacturers and making them reliable again and again and again. So they all seem like they're the same, but if you really tear them apart and you look at all the various components from all over the planet to make that work at tens and hundreds of millions scale, is just incredibly impressive just on the manufacturing side versus, you know, what we were doing when you have extra air and you can move things around. It's like, okay, you don't really need to worry too much about that stuff. It's a whole another... these are layers of the cake of engineering to manufacturing, and they have to be so well integrated. Now, it used to be there was a gap between, you know, engineering and manufacturing, and there's no way, it's all fully overlapped, even all the way down to the chip packaging level.
That's right. And that's, you know, Apple is at the pinnacle of being able to do that kind of integration and being in control of its supply chain that way. And I'm curious to have you think back to 2001 when Apple was a very, very different company. Its market cap was under $4 billion. So like, you know, rough order of magnitude, one one-thousandth of what it is today. That was a company... retail... yeah, no retail stores. Right. So how do you convince people? Right, right. It's just, you know, and they were a niche product. So, you know, thinking back to kind of the uncertainty at Apple, it was, you know, it was struggling financially. How did that uncertainty affect your product development process? Was there stuff that you had to cut? Stuff that you had to just be really ruthless in order to ship it?
Yeah. To get the original iPod out in 2001, it was literally... people didn't even want to do it. They wanted to kill the project before it even got started because they're like, 'We don't even have the cash for this.' Not just the cash to design it, but to market it, right? Like, no one's paying attention to us. Why are we even bothering with something that's not even in our wheelhouse? And so there's executives at Apple today, and I will leave them nameless, who didn't want that project to... who wanted to kill it. And so we had to fight the internal issues and the external issues because Sony at the time was the Apple iPhone Air of the time of how they were doing and everything. So we had to fight internally, we had to fight externally and just execute as quickly as we possibly could because we didn't know if we were going to have money.
The whole company would be insolvent because it was literally two quarters away from not having money, depending on whether it was a good Mac quarter or a bad Mac quarter. And it was highly in debt. So, and I had enough at that point... I had 10 years of disasters under my belt of companies that went under, bad projects, and big political nightmares where they got cancelled, my projects. So I was like, we just got to keep our heads down and ship, and that was it, you know. And I wouldn't say cut corners, but it was how fast can we de-risk and get this thing out the door? Like I said, with air and extra air and all these different things, like, you know, let's show the world what it can be, even though it isn't what we would ever consider shipping in this day and age.
Right, right. If you think back to launching, you know, I'm a marketer. I'm curious about this. Did you have to market the iPod? In retrospect, we think of it as one of these like immediate successes, just like no question. Of course, the iPod, you know, transformed the world of consumer electronics and how you experience computers and everything else. But when you released it, did you actually have to market it?
Oh, it needed to be marketed, but it didn't just need to be marketed. It needed to be fixed three times. And that's what I talk about in the book. Almost any new 1.0 product has to be revised three times before it's actually a success. And I've seen that over and over and over. You know, even ChatGPT has had to refine it till it got to be something useful. And even if it doesn't... it's not just hardware, it's hardware or software or both. And so we had to... it was only marketed to Mac users at the beginning. And, you know, there was a big internal fight about should it be on the PC or not. You know, should the PC be involved in iTunes and the iPod and all those things. And it wasn't until the third version did we get, yes, it needs to work on a PC. Yes, it needs to have a music store that you just didn't... you rip your music off a CD, or, you know, many of your audience probably don't even know what I'm talking about because it was so long ago.
Yeah. When music wasn't just a notional, purely digital, ephemeral thing on the internet.
Exactly. Now it's like whatever. You don't... it's just a snap. And so it took a lot of marketing, but it also took a lot of product revision and listening to the customer because the first one didn't sell that well. The first one, you know, Apple had less than 1% market share. I think we sold maybe 200,000 of the first one, maybe. And it just kind of... first quarter, and it was all the Mac aficionados, and it went dead because you saturated... nobody else had a Mac. Right. And it was very expensive, you know, and all those things. And the second one, we got the product better, but not great. And then we got the PC connectivity, iTunes, music store, all of those things, and made the product with a click wheel, all that stuff. And everything came together. And then it went... it just shot up. But we had to go three years to finally get it to go. It wasn't the first in the market. It wasn't the first year or even the second year that it was going. And then it was combined with how much are we going to invest in marketing? Right. So you're like, ah, we're going to take the risk. Are we going to spend this much cash on marketing? Are we... you know, and so you don't just blow it out the door with your first one. You're going to, you know, you also have to figure out how to speak to the audience and which audience it is.
And so the initial audience was really the diehard Mac fans, and you saturated them pretty quickly because that wasn't a big group at the time. And then, do you remember what year the famous ads with like the silhouettes with the white earbuds came out?
Yeah, that was 2003, 2004.
Okay. So that was as it was going mainstream.
Yeah, it was as it was going mainstream, and then we started seeing the uptake, and then we started seeing... well, there was no real social media at the time, but there was this kind of popular media, you know, that was on TV programs and things of that nature, and people started... you see them on news programs, and you started seeing them on TV shows and things of that nature and popular culture, and then songs were starting to integrate it, right? And you started seeing it with sports players and celebrities, and then it was like, whoa, now it's something, right? And that then gave us the wind to get to the iPhone.
I was a very early iPod customer myself. I had a second generation when I was in college, but I didn't have a Mac, and so I had to buy a FireWire card for my PC so that I could connect it, and then also use some piece of freeware, probably full of viruses, to load music onto it from my PC. But I was dedicated, and it was a word-of-mouth thing as you're describing. I mean, there really aren't that many things that you hear about through just word of mouth. But I remember my first friend showing up after Christmas break with an iPod, and it was just like getting passed around, and people were fascinated by it.
It was viral marketing before such a thing moved online.
And then, you know, what really was the next step above that was the marketing plus the iPod mini, which was the first colored ones, you know, and then the colors matched that, you know, where the ads matched the colors and all that stuff. And it was always small and it's tiny. It wasn't a bigger device, and it's like, oh, it's my cute little thing. Makes it accessible to a lot more people.
Yeah. And it was also a lot cheaper, too.
So, moving forward then, you know, you worked on 18 generations of iPod, is that right?
Yeah. 18. Feels like a whole other life.
And then, by the time you finished up at Apple, I mean, it was much closer to the giant, you know, well-resourced company that it is today. And then you left and you founded Nest, and you started building hardware again outside of Apple's, you know, hardware infrastructure, product development infrastructure. What surprised you about scaling a new piece of hardware, you know, with a decade of iPod experience inside Apple and going back to doing it from scratch without the infrastructure?
The funny thing is, we didn't have as much infrastructure, you might think, at Apple as you might have thought. And that infrastructure was built for Macs. It wasn't built for small handheld kind of devices. So we had to build a lot of that. Like, we couldn't even get small capacitors and resistors on the board because we never at Apple had made those tiny kind of components before, really, or made products with those components or those types of screens and all that stuff. We had to kind of create. Now, there was reliability labs and those kinds of things that we were able to use, but at the end of the day, we had to create it and then recreate it at Nest. So we kind of knew what to do, and we also knew we had taught a bunch of the third-party contract manufacturers how to do it while at Apple, and we got to go work with them, like, you know, some of those manufacturers again and get to reuse it. So it wasn't as hard from an engineering perspective. It was hard because, you know, we had to do it on a shoestring budget and leather stuff and make it feel and be as high quality as that kind of product. I think for me and a lot of the team, it was really the marketing, because most of us had... almost all of us had come from the engineering, mechanical, operations, and it was, okay, how are we going to do the marketing? And I had learned through osmosis a little bit of that, and that I had to really apply myself to that end of the thing and really think it through and build it in a way that you didn't... and this is what you didn't have, which was millions and millions and millions of dollars of marketing budget like that, and a brand that you could sit on top of, right? So that was the real... the product was obviously challenging, don't get me wrong. But that was like the new learning for me and a lot of part of the team is how to do Apple quality at a much lower scale from a marketing perspective, a branding perspective, all those things. And really, the product had to earn... you know, our brand didn't just sell it. We had to make an incredible product that people then learned about our brand through.
Well, and good marketing begins with good product development and good product vision. You know, Nest went into a neglected market. Like, no one thought that thermostats were sexy, that they were worthy of the kind of technological and design invention that you brought to it. Where did you see that opportunity, and then what kind of challenge did that present for your marketing when you actually delivered it?
Well, it was a wild idea that... it was a pain point for me for many years because I had a remote house in the mountains and I wanted to control the heating and cooling. And I also didn't want to spend a lot of money wasting on heat or cooling when I wasn't there. And so it was like, okay, how do you make a product to do that? So I was waiting for a product on the market, and I was also cobbling together, kind of maker style, trying to put something together, and nothing was working. And that was over 10 years of trying and trying and trying, and I was just like, man, nobody's figured this out. And then the iPhone was just about ready to exist, so it was in the back of my head, I'm like, wait a second, I could do this, I know how to do this. Like, okay, we have to... you know, there's a few components we didn't know how to do, but for the most part it was a solvable thing. What I didn't know was AI, and we put AI in the product, and so we had to get AI experts for AI at the time in 2010, 2011, and we were one of the first products that ever had AI in it, especially consumer products, and that was, you know, a million-unit seller, you know, many millions of unit seller. And so the pain point was this remote connectivity, and then it was like, and this kind of make it intelligent, make it so it's not just you up or down from anywhere, but make sure that it can, you know, sense when you're there and not, and really try to imbue all of these kinds of characteristics that we're now seeing in products back in 2010, 2011, trying to make them very smart. And smart's an overused word, but ours truly were. And then we did the same thing with smoke detectors and cameras and other things, right? And but we were just on the forefront of that, and it was all just pain points. It was pain points, as you said, of neglected products that cost money, needed to save lives, needed to protect, you know, assets and things. And so we just went from, you know, homeowners really need this stuff to let's make something better because no one had touched these things in 20, 30 years using the latest technologies and connectivity that we knew about from the products that we had already previously designed and built.
If I'm imagining introducing a new piece of intelligent, you know, home hardware in 2010, I can think of a lot of challenges related to like distribution. You know, that the Home Depots and Lowe's of the world are not... they don't even... it doesn't even occur to them that they're on the lookout for the next great, you know, technologically advanced thermostat to bring into the stores. Did you have to go and completely learn the world of home goods distribution and, you know, getting into hardware stores and stuff like that? Did anything surprise you?
Well, first we had to learn about thermostats in general and the world of HVAC, right? And I was like, we're going to hire some experts out of the HVAC industry. Well, the experts were either dead or they were so encrusted in that world that they couldn't think any differently. So we had to just go and figure it out ourselves by reverse engineering these systems and going out to all these homes with multimeters and oscilloscopes and ammeters and going, 'Okay, what are we doing?' You know, and... 24 volt AC. I wish it was that simple. The amount of noise those products are like a disaster from an electronic design perspective. They're mechanical products that some mechanical guy hacked together some, you know, electronics to sort of work. Like, and so we had to go learn that. Then we had to learn the HVAC industry and installers, and then understand installers were not going to be a route to market, and then going to home... not Home Depot, but going to Best Buy, who'd never done a thermostat in their life. Right, right. And then Home Depot was like, what electronic, what? They were like, 'We just sell these things.' Like, Home Depot. Now, if you go into a Home Depot, if you go into a Best Buy, they have full home automation, like retail racks and all these, you know, try-it-before-you-buy-it kind of things, and you're like... it had none of that. And it reminds me of when we were doing, you know, PDAs in the late '90s when it was like, is this a calculator, a computer? Where does this thing go? You know, where does an iPod go? Does it go with the stereo equipment? Does it go with the computer? So we're always, you know, if you're pushing retail to come up with a new category, you might be doing the right thing.
Yeah. It's a very conservative industry. They're just doing what's always worked, and that probably means there's opportunity for you as an entrepreneur, right? Exactly, exactly. You mentioned, you know, putting AI in the Nest thermostat, and for listeners who haven't used one of these thermostats, I think what you're talking about is the fact that it learns from your day-to-day adjustments, you know, bringing the temperature up around the time you're waking up and then turning it down maybe when you go to work, turning it back up in the evening, and so on. And it recognizes those patterns and starts to make a schedule for you. Right.
Correct, correct. And we had to call it the Learning Thermostat because we couldn't call it the AI Thermostat because people go... now if you don't have AI in your name of your company or your product, you're like, 'I don't want that, I want the latest and greatest.' 15 years ago it was like, 'That's scary, right? I don't want that controlling my home.' So we had to come up with new language, had to totally bury it, explain it in real simple details of how it works, and, you know, kind of whitewash it away like, yeah, don't worry how it works, it just works, kind of a thing. And, you know, it was quite a learning experience because I had never done AI at that point, and I learned a lot. And what's very interesting is now with AI and knowing what I know then, I'm like, oh, how did you train it? What's the training data? What's... because we had to go through all of those things, and how do you make sure it doesn't do something weird like hallucinations like today? And so it's very interesting that those same rules still apply and will continue to apply regardless of whether it's an LLM, an LM, you know, whatever the AI is. It's really about training data sets and everything.
Well, so tell us a little bit about that because as an investor now, you see a lot of hardware startups, must be seeing a lot of AI in hardware startups. So tell us about the challenges you faced bringing AI into the Nest thermostat, and do you see everyone making the same mistakes or trying the same things today at a fundamental level?
Yeah, I think for us it was, you know, we were just like, could we even make it work? Like now you're like, sure, of course it'll work, you can just simulate and... like back then we're like, can we get enough data? You know, here's the one thing that is very similar today that was very similar today as it was back then, which is when you put AI in a product, how quickly does it learn you and your habits to be able to work and give you benefits right away? And so when you have small products with small memory... and we're not talking about connected to the net with these large LLMs with, you know, gigawatts of processing. We're talking this had to work on a battery-operated thing in 2010 with just the processing it had and the memory it just had, and it still had to be cheap, right? And so when you think about that, we're still... everybody today is still going, okay, what's my cost per token? And how quickly can I get done? How can I get my power down? How can we get all of these things down so that it can serve a lot of people, whether it's on your device like Apple's trying to do or Samsung, or in the cloud? Everybody's still trying to compact that. And, you know, it's when we started with something that was so constrained, we had to try to... we had to make it work in that highly constrained environment. And there were other techniques even back then that we couldn't use because they were too powerful or consumed too much energy or too much memory that we couldn't even use. So we had to figure out one, AI, and two, can we even make it work in this incredibly constrained environment, which is way more constrained than we even have on the smallest devices today in the market.
Right, right. What are you seeing among the hardware startups that are trying to bring this to life, and the bigger companies? Like, what mistakes do you think they're making in trying to bring AI into the physical world through hardware?
I think it's, you know, I think it's here are lots of capabilities and a lot of things, and you can start with the chatbots first. You know, you don't know what they're good at and what they're not good at. You have to take a swing at it and go, 'Oh, I think it works.' You know, and then you give it one math problem and it solves, you know, a math olympiad question. And then you give it a simple thing like count the number of Rs in the word, and they're like, 'Can't even do that.' Right. So you're like, it's so imbalanced and so jagged, and you don't know what it's good and what's not. And whether it's really big like in a server or it's down to, you know, small products, what people don't understand is AI is a technology. It's not a product, and even a chatbot is not a product. It's not been productized. So if you want to try to use that chatbot in a medical application, in a finance application, in some kind of critical thing, systems control, like let's say thermostat control or energy controller, you got to really understand what it's trained on. You have to understand all of the permutations of that to guarantee it will work in these critical situations. Otherwise, you can't rely on it. And you're going to always need the human there, right, to check everything, which is fine, but you can't sell it like, oh, it's this answer engine, but yet you don't know if it's answering it properly. And if it's in critical stuff now, the bar's raised that much higher. And it doesn't matter, like I said, if it's a huge system or if it's a tiny system, it's got to work in these critical situations. And too many times people are like, it does everything, but it really doesn't do any one thing so well that you can rely on it. Like we have agents, but you're like, well, the agent didn't work for me this day and it worked yesterday and it sometimes didn't. You'd fire anybody who did that if they were on your team. You know, you got to rely on them. And so I think we're still in a very, very early stage where... and I don't care how impressive the results are of Claude or, you know, OpenAI or whatever it is, if it still has this jagged thing and you can't rely on it, then it tells me it hasn't been productized so you really know what you can rely on it for.
It's interesting, it's a great way of putting it. One of the things that I really think is brilliant about OpenAI's marketing around ChatGPT is that they've typically shown it being used in lower-stakes settings. You know, famously sort of Studio Ghibli-fying your images, or, you know, asking about a restaurant reservation, or kind of like lightweight things like that. They're not showing it controlling a power plant or, you know.
Yeah. Then they'll say it wins the Math Olympiad, and then it does this and it does, you know, it... and so again it's like, see how impressive it is, but then when you go to use it, it's kind of a letdown. Like, still try to get a photo out of it that you, you know, photorealistic thing, and you give it all these prompts, and it still breaks. So we're still at the very, very earliest days, and frankly, I think there's a lot of technology that still needs to be produced. It's not just scaling that's going to make AI work. It's going to be a lot of better training data, and also models, not just LLM models, but other things, neurosemantic models, reinforcement models, other things to really make it so that you can rely on it. And it has to be scoped down because once it's scoped down, it's understood, then you can also price it right so it can be attractive for people to use it. Right now, you got billions, almost trillions of dollars going into this, and everybody's forward-pricing everything, but no one's making any money because it's just not in the right... it doesn't have the right parameters yet. Everybody likes it in some form, but when you say you're going to rely on it and it's going to cost you $2,000 a month to use because it really works, people be like, 'Well, wait a second. I don't know if I want that.'
Yeah, yeah. And I think people forget in the physical world or with hardware products, with, you know, things that act on physical things, specialization is the way to success, and eventually you can become a little bit more generalist. But, you know, you could argue that the smartphone is the, you know, the epitome of general-purpose hardware. It can do all of this stuff. It's always in your pocket. It's connected to the internet. But it's easy to forget how long it took to get there, right? The first iPhone was just a handful of functions, no App Store. It took years before the iPhone was something that could like replace your laptop on a short airplane trip, or, you know, or was something that you wouldn't just like put down and move to your computer to write a reply to an email.
Exactly. And it's still, in some ways, even an iPad today, a tablet, is still that people are like, well, when do I use my tablet, when my computer? And they're like, okay, I'll just default to the computer. So we're still trying to get there. And, you know, people like, look at the adoption curves of AI versus everything else. You know what? It's riding on the shoulders of giants. You know, I remember when we didn't have the internet, and then we didn't have, you know, broadband, then we didn't have Wi-Fi, and then we didn't have mobile, and we didn't have social, and then we didn't have, you know, incredible processing and central servers. And so when you finally built out this entire network and everything end-to-end is like starting to really work, and then you go, 'Oh, AI shows up and then it goes everywhere and everyone's using it.' Well, sure, duh. When you've made it frictionless because you've set up the whole planet to allow it to happen, of course it's going to go that way. It doesn't mean everyone's going to pay for it in the long run. It doesn't mean... because you still have to show incredible value because they're going to have to price it at what it really costs to deliver, and they're not really doing that. Like, they're giving it to governments for 49 cents and what have you. And so there's still reckoning, and there's still a lot of stuff going on. Like you said, just because the infrastructure was there to allow it to flow everywhere doesn't mean it's amazing for lots of people. It can do lots of amazing things, but for who, for whom, and how often, and can you rely on it?
Right. So in that stack of technologies that were the enablers that you just listed, that, you know, have made the modern world what it is, what do you think is the next one? Is there something you see that's emerging now where you're like, ah, that's going to be, you know, at the foundation of the next generation of great products, you know?
I think we are at the point where, you know, you hear of the super apps, and there's like everything under the sun in the super app, or you go to your neobank like a Revolut or Robinhood or something like that, and they have tons and tons of features in it. And these apps, and, you know, they're like the Adobe apps or Canva or Figma, and you're using them mobile or whatever. I see... and there's just a plethora of palettes of tools and everything else. I see that inverting. I see these palettes of tools going away, and you're going to have an agent that you're going to... who understands the software really well and say, what was it you want to try to accomplish today, and then start to surface things. And you're going to start to see these interfaces be much more dynamic than a static thing. You know, you look at your, you know, your Outlook or your mail app, and it's always the same things, and 60% of the screen real estate is all tool palettes and stuff, and you're like, I got a content window like this. You know, it's going to learn over time, just like the Nest Learning Thermostat did, is like, you don't use any of that stuff. It's going to go away, and it's going to surface it when you need it, or you can...
Ask it, 'Hey, I'd like to do this. How do I do it?' So that it will... so these tools will get smarter to work with you, and it's not just this barrage of tabs and buttons and things of that nature. I think that is a logical next step that so many apps that are already highly used tools by many, many professionals are going to get simpler so that more people can adopt them and get superpowers. I think that's one. It's not going to just be like, 'Hey, you go off and make me, you know, AI, go make me all of these things and then I'm going to just use it like that.' You're going to want some kind of, you know, sure, some starter thing that the AI can do, but then you're going to want to have the tool pallet and everything to work with in concert with it to really bring to life what is in your imagination, because prompting is not enough. Just texting, it's like the old command line days, you know, that's only going to get you so far.
Yeah, it's one of my favorite UIs is Photoshop because it kind of displays very clearly the, you know, the evolution of the software over like 30 or 40 years, right? And you have these menus that are hundreds of options once you really drill into them, and it's like save as (new), save as (legacy), save as (you know, something else). And it's like everyone has built their workflows on this since the mid-1980s, and the software still supports all of those workflows, but it's forbidding to anyone who wants to, you know, use the software to be creative and has never done it before. The learning curve is so steep. So what you're describing would make really powerful tools like that accessible to a much wider variety of people.
Exactly. And just so you know, Photoshop wasn't until '91. I know because Photoshop was built — Tom Null built it in the lab that I worked at at General — at University of Michigan, and I had the little thing. It used to be a photo booth. It was called Photoshop. But my — and I hope that Lumafield is also going to be doing something like that, because you're going to have such incredible power in these scans and everything that your interface is going to be able to be much more morphing like that to give people, like, 'Oh, I want to do this kind of analysis and this kind of thing,' whether it's in, you know, real-time flow on a manufacturing line or if it's in one-time testing, destructive testing, these kinds of things. So I'm looking forward to seeing how, you know, the innovative team at Lumafield, you know, starts to adopt these things not just at the analysis level, but at the tool level to give people superpowers.
Yeah, that's part of our vision. I mean, we've implemented an agent called Atlas that's inside our analysis software, Voyager. It lets you, you know, talk to the agent and talk about the analysis you want to run and automate things. We've seen some great uptake on it.
That's great. Well, I'm looking forward to the next versions. You know, you guys are smart, so it's going to be wild to see what the mere mortal be able to do with it.
All right. And with that, I'll wrap up by asking you the question that we ask everyone who comes on this podcast. Can you tell us about a go-no-go moment that you've encountered in your work? How did you face it?
A go-no-go moment. It's the title of our podcast.
Yeah, I think, you know, yeah, it is the title of the podcast.
I think it was really about whether or not I should do Nest. I had already done so many things. I was like, do I really want to take on this challenge? And I put a little bit about this in the book, which was, do I really want to spend the next, you know, 10 years of my life building this thing? Because I know how hard these things are to build from scratch. And I was like — and I had already 20 years, over 20 years in the business. And I'm like, 'Oh, I know how painful this is going to be.' I wasn't just like a 20-something going, 'It's all gonna be easy.' So I had the scars, and what was really interesting was when I would talk about the problem, the solution — that with whoever it was I was talking with who, you know, understood it well enough, they would just go, 'How can I help?' And so the path kind of — while it was very cloudy and dark in ways because I had so much previous experience, it opened up and the light shone when, like, 'Oh, I'll give you money for that. Oh, I want to join your team. Oh, we'll help you with the chips. Oh, we can do this.' Oh, and it was just like — and it was just like, oh, it's go, because it was like I was like, no, no, no. I'm not going to be a glutton for punishment again. No, no, no. You know, and then it was like, 'Oh, I have to do it.'
Yeah. The ecosystem called you back in.
Yeah. I love it. Well, Tony, it's been such a pleasure to talk with you. You've always been a great supporter and we really appreciate you coming on the show today, and I encourage all of our listeners to read Build, Tony's incredible book about great product development and great companies.
Oh, hey, thanks John. Thanks a lot, and looking forward to seeing this episode and future episodes.
All right, thanks so much Tony.
Welcome back. I'm John Bruner
For our next segment here on Go No Go, we're going to talk about some of the news headlines that we've been reading about manufacturing and product development and technology in general. Alex, what do you have for us? The first story I wanted to touch on was an article by the New York Times about how Trump's tariffs are damaging our largest source of screws.
And so I think most Americans are familiar that a lot of computer chips come out of Taiwan. But Taiwan has actually been the number one source of screws and metal fasteners to the US for the past 30 years.
These are objects that you don't think about in terms of the supply chain. They just are.
Exactly. They're something you really take for granted because they're so essential, right? And so far down in that supply chain. But with the 50% tariffs on steel and aluminum, now these Taiwanese producers are facing a lot of pressure. You know, their margins have already been pretty thin. They're seeing competition from China. And so, while the bigger companies are able to move to other countries within APAC, a lot of the smaller ones, unfortunately, are having to close.
And how's this affecting American manufacturers?
I think it'll be interesting to see how this plays out and its impact for American manufacturing. Something the article points out is that, you know, TSMC even is having issues hiring like high-wage laborers for their fab in Arizona. So, you can't really imagine that a lot of Americans are going to want to work in like a screw factory.
Yeah, this whole thing is a really interesting illustration of how disruptive tariffs can be. You know, if you think about where US manufacturing is really competitive, it's in very high-value complex products like airplanes and surgical robots, and things like airplanes and surgical robots have a lot of screws in them. So, if you just sort of apply blanket tariffs to a lot of the upstream inputs in these products that American manufacturers make, it can be quite disruptive. The same is true of applying tariffs to base metals like steel and aluminum. You know, we would all kind of find it thrilling if there were cities in the US that could go back to looking like Pittsburgh in 1955 with like smokestacks and huge steel mills and so on. But it is advantageous to the US automotive industry among others to have access to inexpensive metals, and that is what, you know, helps them thrive as assembly operations and integration operations in the US. So this is one of many kind of inputs to American manufacturing that's getting reordered under the tariff program.
Yeah. All of those costs really stack up. And I think what you say too about like higher-value manufacturing really being more attractive is even true in Taiwan. A lot of these screw manufacturers talk about how they have to compete with the computer chip industry for government support and for labor, because of course people maybe prefer to be in the area that's more lucrative and more flashy. So it's a real challenge even there.
It's something you hear factory managers here talk about a lot. It's also really hard to hire employees at manufacturing operations at factories in the US for exactly that reason. You can make pretty good money working in an Amazon warehouse or in a coffee shop. And you know, there's a generational difference in the appeal of factory work. A lot of younger people have grown up with their parents saying like, 'I hope you don't have to work in a factory.' But those can be really good jobs and can be quite lucrative if you have the right skill set for it. Speaking of tariffs, you found something interesting about Lululemon and how they're distributing their products and how this is also being disrupted.
Something I was really fascinated by in Lululemon's latest earnings report, and they also covered this in Supply Chain Dive, is that the de minimis exemption and the end of that has really materially impacted their business. So apparently about two-thirds of Lululemon's e-commerce orders in the US are shipped through Canada using the de minimis exemption.
Interesting. You think about the de minimis exemption being used mostly by companies that are shipping straight from China to consumers, but it seems like Lululemon found a different way of organizing its distribution.
Exactly. They had seven distribution centers, four of which were in Canada, and so they could ship these individual consumer orders across the border using that under-$800 exception and save a lot of money on taxes. So now that they're having to pay those taxes while they figure out how to reallocate their distribution centers, it's going to have a real negative impact on their margins.