“Our Goal Is to Build an Electrical Engineer.” (Davide Asnaghi, Co-Founder & CEO of Diode)

16:50Davide Asnaghi:And this is changing in many ways. But so many of my friends were just for standard deviation above average, like knowledge in electronics. People that had been building electronics for like years and, you know, they were immersed in this culture. And I think that that left a profound impression that this not only can be done, but also there are people that are immensely more knowledgeable than I could ever hope to be at that point in my life. And they are incredibly graceful at teaching this kind of stuff. There's this person specifically, Alex Wong, him and I worked on a couple of papers on a robotic hand for prosthetic devices.

17:32Davide Asnaghi:and like he was 19 at the time. And to this day, he's probably like at that point in time of his life, one of the best electronic engineers I've ever met. And I've met a lot. So it was just like a really interesting time to absorb from people like this.

17:47The Generalist:What was it that had drawn you out there? Was it to do specific types of research? Was it just to get that sort of different experience?

17:54Davide Asnaghi:It was really like, I truly enjoyed my time in the United States. and I knew at that time that electronics was a very important part of my life. And it seemed wise to sample what the rest of the world has to offer in a domain that I truly care about. Like arguably Shenzhen to this day is one of probably the most lively hardware hotspot in the world. And you want to be close to this type of centers to at least learn. And there's a lot of learning, like also negative learning that happened. Like, for example, starting a company in Hong Kong and China has drawbacks. IP is more complicated. The legal system is slightly different than the United States.

18:39And these are all things that are very hard to learn if you don't live immersed in the culture.

18:45Davide Asnaghi:Like, I was an exchange student in Hong Kong, and a lot of the exchange students are there for like a semester, and they effectively create their own social bubble. But because I was like a full year, the idea was I was actually like hanging out with people from either mainland China or like native Hong Kong. And the cultural difference was just staggering, like in many ways. And it was a wonderful experience to learn that.

19:10The Generalist:What are some of the things that Shenzhen does much better than Silicon Valley in your experience? It's very hard to say better.

19:18Davide Asnaghi:I think that like Shenzhen is a fundamentally slightly different beast. Like in Silicon Valley, you had these wonderful network effects because if what is limiting your productivity is a keyboard and a computer, the network is largely composed of social relationships. And social proof, the ability to know that there is vouched talent at the fingertips has been an incredible engine of Silicon Valley. what I see in Shenzhen is a radically different thing where it's not necessarily just about social relationships. It's actually about the community as a whole dividing complex manufacturing processes into like a divide and conquer approach.

20:03Davide Asnaghi:And then each of the individual manufacturer like brewing manufacturing expertise and then being able to hand it off to the next step. And so the knowledge of how to make things permeates and percolates through the various layers of the stack

20:20The Generalist:in a way that Silicon Valley, I don't think,

20:24Davide Asnaghi:has been focused on achieving so far. It's not that it couldn't. It's just that it didn't seem like the right problem to be solving because a lot of the other problems were how do we improve productivity by unlocking new software, like new computing paradigms. And I think both of these type of realities have been very successful. And I think that there's a lot to learn from both. Like, I think that it's very hard to replicate what Shenja has, but not impossible. And in fact, I am seeing a lot of what I used to see with my friends and like people that are starting companies in the manufacturing space.

20:55The Generalist:To put a finer point on that, like what sort of things are you seeing? Like, is it cultural things? Is it sort of operational things?

21:02Davide Asnaghi:I meet somebody for coffee in Silicon Valley. And then four weeks later, they need a board made fast. And this is not like the exact service that we provide. Like what we provide is we will design and manufacture. And because we can co-design these two things, we are massively more efficient at the manufacturing. But if you're my friend, I'm just going to do this for you. And if you are happy with this type of service, I'm happy to like basically expand this to a proper business relationship. Or I have a customer that needs motors. And I know David Hansen in Silicon Valley, who's building a motor company.

21:39Davide Asnaghi:Hey, like you guys should talk. And so because the stack is very wide, like printed circuit boards are very hard to build holistically, but also incredibly narrow in the stack of what it takes to actually build a full product. I get so much more reward from being able to introduce my customers to the people that will fix their pain points. And everybody wins. And this I am seeing more and more and more and more. And it's quite wonderful to see.

22:06The Generalist:You come back to the States and you sort of spend quite a lot of time at Apple in their sort of special projects division. I'm sure there's a lot of things you can't talk about. But what did you learn from sort of being part of, you know, the best hardware company in the world, arguably?

22:23Davide Asnaghi:So special projects is a very unique part of Apple. So I don't want to pretend I understand Apple as a whole. I had a very peculiar experience. And the experience was largely determined by working with some of the best people in the world, whose relationship I cherish to these days. My manager at Apple was a wonderful person that had a ton to teach from an interpersonal relationship perspective. The thing that was missing for me from that experience is that you are, like, if you work at Apple, you own a very, very narrow part of the stack, but you do own it incredibly deep. And that is, like, very good for honing a very specific skill set.

23:04Davide Asnaghi:But the moment you want to own a little bit more, it's very hard. And for very good reason. Like, there were folks that were owning a larger portion of the stack with lower depth. And these people had done it all. Like these were some of the best engineers in the world that had like a system engineering background and were responsible for requirements and capturing. And I think that like apart from like many technical things I learned and it was a wonderful experience. The thing that I learned the most is that like electronics was not a solved problem, even for like a company in the U.S. that is largely the best at doing this.

23:42Davide Asnaghi:Like Apple has clearly solved electronics at scale for production, but our prototyping cycles and the speed at which we could iterate was a fraction of what I was used to in Shenzhen. And I think that that was a fairly shocking realization. Like it felt that this is like a systemic problem, not an individual company problem or an issue that comes from lack of resources or lack of knowledge. I think that that was the key learning I made.

24:11The Generalist:Psychologically, you know, people who become founders, I think, often struggle to be in as structured or as large an environment for such an extended period of time.

24:22Davide Asnaghi:I think that you might be spot on with your read.

24:24The Generalist:I was there for about two years. There was a part of you that was itching to take on some of that more, some of that responsibility.

24:31Davide Asnaghi:I think so. And in fact, like when I left Apple on paper, that was a questionable decision. Like I left Apple for a much smaller company. I left Apple in the middle of my visa proceedings. And this was not out of frustration. Like it was a wonderful place to work. It's very hard to work in an environment that does not reward you basically killing yourself over your job. And like, this is not good for everybody, right? Like you don't want everybody having to like work to the max. But in this specific period of my life, I generated such a high amount of personal enjoyment from like dedicating myself 100 % to like this one thing.

25:15Davide Asnaghi:And Apple was a great place to be, but it's not structured so that you like you basically have a tapered off reward the more work you put in.

25:23The Generalist:Like you can't stop working at a certain point and it's like mostly fine and everybody will be happy.

25:27Davide Asnaghi:And it just felt like I wanted to be in an environment where I could put as much of myself as I could into the craft.

25:33The Generalist:Have you always been that way that you want to sort of kill yourself for things that you're passionate about?

25:38Davide Asnaghi:My fiance will be very mad that I say kill myself. Let's say being very passionate. Yes, that's a better language. It just feels that I have an incredible privilege of loving something that is very, like it's perceived as valuable in the modern world. Like if I were like a musician, for example, it would be much harder for me to just dedicate 100 % to my craft and like also pay bills, right? But like because the things that I love are very, very aligned with what the modern world deems like financially rewarding from like as an engineer, you can make a living, no problem. I just feel like an obligation of basically like this is the few set of talents that I have.

26:21I want to be able to pour myself into it.

26:24Davide Asnaghi:I also like find it super fun. Like, this is just like, I would do it for fun anyway. Like, I studied biomedical engineering in my undergrad because I think that biology and physiology are like a completely different tier of complexity than electronics or even silicon. It's just like a language we don't understand yet. And I wanted to like learn more about it. But my personal things were in electronics even during undergrad. Like, I would do it even if I weren't paid.

26:49The Generalist:So it feels like the optimal use of my time. I just get a ton of like joy out of it. That's amazing. You mentioned, you know, you go to this smaller startup, Chromatic, and it's there that you meet your future co-founder, Lenny. Yeah. What was it about Lenny that you saw something that you felt like maybe this was a larger partnership or someone you hope to work with for hopefully many years to come?

27:15Davide Asnaghi:So funny enough, I actually met Lenny on my very first job out of college. So I started working for this company called Butterfly Network. Wonderful, wonderful people. incredible product. It was a custom silicone ASIC that would allow you to do ultrasound imaging in the space of a handheld ultrasound machine. This is revolutionary technology. Very, very fun to work on. Lots of technical challenges. Very smart people. And I had a master's degree at that point. I did my undergrad in three years and my master's at Berkeley. And then for some reason, I think that that counted as work experience. And so they assigned an intern to me, even though I had no reason to need or deserve an intern.

27:59Davide Asnaghi:And that person was Lenny. Wow, wow. And it was very clear that he was way smarter than I am, incredibly capable. He had done an internship the year prior on the iOS team, which is very different than hardware. And he was incredibly well-versed in machine learning, front-end, just like a true full stack engineer. Like while I go like a kind of embedded software down to the metal, he goes embedded software up to like the cloud. And so it felt like a super fun set of combination. Like it's a very complementary technical skill set, at least like in software plus hardware. And I just enjoyed working with him.

28:38Davide Asnaghi:And then when I joined Chromatic, everybody knew Lenny and liked Lenny. And we love it very hard for Lenny to work with us.

28:46The Generalist:And so purely coincidental that you ended up at Chromatic at the same time after that experience at Butterfly.

28:52Davide Asnaghi:I think I was part of why Lenny came and joined. I think that there were two people, one person that also worked with Lenny and we were like, yeah, he needs to join. He should join Chromatic. And Chromatic is also probably the most fun I've ever had as an employee. It was just a wonderful company to learn. Also, incredible founders. We are still very close to these days. We learn a lot from them. And it's quite humbling to see how many things they got just right. And as I build a company, it's just like, how did they do it? What can I learn from them, from their experience? Incredible people. Very, very grateful to have met them.

29:33The Generalist:And it was acromatic that you sort of have this sort of first inkling of this need that becomes diode. How did that sort of come to fruition or come into your frame of vision?

29:46Davide Asnaghi:I think that it was a culmination of observing the same problem over and over and over at different, like completely different scale of companies. Like Butterfly was a pre-IPO company. Apple, obviously, like largest company in the world by capitalization at that time. And then Chromatic, like a fairly small startup company. And every time that we had to spin a circuit board, But despite that not being the most complicated thing that we were doing as a team, it would just go wrong. The manufacturing would take weeks more than it was supposed to take. The boards would come back and have a yield rate that was abysmal, like less than 50%.

30:28Davide Asnaghi:And you would ask yourself, is the problem in the design? Is the problem in the manufacturing? What's going on? And it felt just wrong. And so I started diving deep into this. And it turns out that a large reason why this is the case is that a lot of smart people have poured their hearts and souls into making silicon design more efficient.

30:51The Generalist:Because the cost of getting silicon wrong is in the millions of dollars.

30:56Davide Asnaghi:The cost of getting software wrong is virtually zero. You just push a patch. there's some caveat. It's like obviously if you get something wrong on a security that can be very strong. But let's say during prototyping you can basically just fix it as you go. And PCBs or circuit boards are just in the middle. They're like this weird middle child that hasn't been loved enough. And a wrong spin may cost you between$10 ,000 and$80 ,000 which is bad but it's not destroying your company bad especially with US companies. And so a lot of very standard problems that you'd think like, sorry, but like this should be solved.

31:33Davide Asnaghi:They haven't been solved. And I think that it was incredibly apparent when we taped out this incredibly beautiful piece of silicon that a way smarter engineer than I like put together. And like this guy got it right first time. Like he designed the entire silicon and the silicon just worked. I was responsible for like a small part and I was wetting bullets. I was like, if this goes wrong, like the entire silicon is not going to work. I was responsible for the bootloader in the device. and then the silicone comes back and the circuit board that it was on was like full of mistakes and it was like through nobody's fault it was not like that you could say oh the person that designed the circuit board made a mistake no like the tools were not helping the manufacturing was not helping like we had to diagnose like yield problem like all these like little things that slowed down the most beautiful moment of you get back your chip and you know that it works and like the chip worked first try I am still amazed that that was the case.

32:30Davide Asnaghi:And it was absolutely wonderful. And it's really a testament to how good the team was and also how good the tools that were designed around that were. But it felt that we deserved better tools around those circuit boards as well.

32:43The Generalist:This episode is brought to you by Brex. If you're a founder, the hardest part isn't the idea. It's scaling fast without getting buried in back office work. That's where Brex comes in. Brex is the intelligent finance platform for founders. With Brex, you get high-limit corporate cards, easy banking, and high-yield treasury, plus a team of AI agents that handle manual finance tasks for you. They take care of things like expenses, all according to your rules, so you can move faster while staying in full control. One in three startups in the U.S. already runs on Brex. You can, too, at brex.com slash Mario.

33:22The Generalist:Just to restate it in the sense that to make sure I understand it, It's like, you know, so much effort, investment, intelligence has gone towards optimizing the silicon because it's so important and expensive to get wrong. And then software has these fast cycles that are sort of forgiving. And so there's, you know, plenty of architecture around that. But the circuit boards are in this weird middle zone where it's not that forgiving, but it's also, you know, not catastrophic to failure. And so it sounds like it's sort of remained almost, I don't know, artisanal ad hoc where there's really not this infrastructure around it.

33:59The Generalist:I guess, one, you know, am I sort of capturing the contours of this correctly? And then two, why has there not been more effort dedicated towards this until now? Because clearly there's been plenty of investment in this ecosystem. system, why has this part just been sort of overlooked beyond that maybe structural issue?

34:20Davide Asnaghi:So I think that you captured it beautifully, but there is one important and huge caveat. This has been solved in Southeast Asia, which was the geography where this mattered the most. Like effectively, Western countries around the 2000s collectively abdicated a part of the stack that they perceived to be, you know, not catastrophic failure, not necessarily my core IP, like the same way as the silicon. They abdicated it to geographies where they could outsource it or offshore it for a cheaper price. I see. And if you look at like a modern circuit board factory in China, there is nothing artisanal about that.

35:04The Generalist:The artisanal nature,

35:05Davide Asnaghi:which I think you capture beautifully, like it's quite beautiful because you can actually build a circuit board by hand, something that like you obviously cannot do with silicon. So it is truly like if you have a good artisan, you can build a fully working circuit board by hand. There's some caveats, of course. But the thing that is more interesting to me is why do we not have like the level of precision and quality control that we have for silicon for circuit boards? And if you see the machines behind me, these are industrial machines. These are machines that are meant to address the gap between artisanal product and mass manufacturer product.

35:43Davide Asnaghi:And it's not true that people have not solved it. But where it hasn't been solved, especially in the US, is a fully autonomous pipeline that takes you from your design to a manufacturer product. Effectively, like the case that the US market has been optimized around is, I have a design, we have iterated on it. Now it's time to build 100 ,000 of them. There's a lot of incredibly competent shops that will do this in the U.S. And this is not true for circuit boards only. It's true for CNC machining, like really any type of manufacturing. The best manufacturing shops stateside thrive in volume because variance is the enemy of yield.

36:25Davide Asnaghi:As soon as you need variance, there's a lot of complicated calculations that you need to do. Like for a PCB, it's like you need to figure out how to place the components. You need to figure out how to place the solder paste. You need to figure out the optimal reflow profile of your oven. You need to figure out if your optical inspection machine is calibrated correctly to spot mistakes, because this is a manufacturing process. There will be mistakes. It's just how do you minimize them? It's not like this is always going to work. And I think that because China started with cheap labor, and then effectively labor progressively became more expensive, they truly invested in wonderful automation.

37:03Davide Asnaghi:Say, I will not trust somebody that tells me, oh, China can do it because the labor is cheap. That ain't true. That's not true anymore. They are incredible at automating things and automating them in a way that makes them cost competitive. And yes, there are a lot of things that are, you know, there's an 80-20 of automation where the 20 % is filled in with cheap labor. But if you look at a modern circuit board like factory, it is incredibly automated. it. And I think that this is very hard to replicate the novel if you don't do something radically different. And our bet is that that radical difference is that you co-design the circuit boards with the assembly line in mind.

37:41Davide Asnaghi:Like if you have these two concepts in your head

37:44The Generalist:at the same time, which is a very painful process to have, it will make design harder to know that

37:50Davide Asnaghi:like if you place a via with an annular rank that is not like to spec of the machines that I'm using to drill the bare boards, it will cause a problem during manufacturing. It is very restricting. But the bet that we are making is that the models don't care. Like you can let the model iterate until the design is as manufacturable as you possibly can. And so this yields a much more like amenable path to full automation design to production.

38:15The Generalist:In a way, it sounds like you're sort of uniting these two pieces of the input and the output. You know, you have the ability to have the, you know, the system understands the inputs you need to make, but also how those get translated into the actual outputs and the sort of capabilities that you have on hand.

38:32Davide Asnaghi:And like, I'll go one step further, which is like, if you look at the cultural stance that like my friends in Hong Kong used to have, every board is designed thinking of, oh, I'm going to make this one sided, even if it costs me like a little bit longer design time, because this way doesn't have to go twice during assembly.

38:53The Generalist:I see.

38:54Davide Asnaghi:I would ask them, like, why? Like, just designing both sides, like, who cares? They're like, no, no, no, no. This will have to, like, go twice during assembly. Like, the yield will go down because you're reflowing the components twice. Like, if I can do it, I'll just do it, like, one-sided. And that, like, cultural mode is just, like, has been diluted in the U.S. because it's so removed, right? In a lot of ways, when you send your design to Southeast Asia, they will tweak it to make it more manufacturable without necessarily telling you, which is quite interesting, in my opinion.

39:24The Generalist:Yeah, that is interesting. What are the companies that sort of dominate the industrial scale production of this in China? Is it sort of like actually a market that's relatively diffuse or are there a few sort of big players?

39:38Davide Asnaghi:I think that there's two clusters. There's Foxconn and Pegatron. Like these companies are huge companies. They don't do just circuit board production. They really actually do the full stack. We would classify them as EMS, electronic manufacturers. These companies are incredibly successful. Like the story of these companies is wonderful. Like Breakneck by Daniel Wang has a lot of like really interesting tidbits

40:01The Generalist:about like how these specific companies were founded and generated.

40:05Davide Asnaghi:And like, I think that these companies are one end of the spectrum. But then you have other like younger companies. There's a lot of mom and hop shops, by the way. Like it's not true that like PCB manufacturing is a monolith in China. And of these, I think that some of the more famous companies in the US are like JLC PCB, for example, which is this Sequoia China backed company. Incredibly interesting company that doesn't do just PCBs anymore, but they started with circuit boards. And if you look at the way that they are automating a lot of the stack, it basically makes the same bets that we are making.

40:46Davide Asnaghi:I don't want to claim ownership of these ideas that we are putting forward. I'm just looking at what's working and basically saying, okay, they do it in this way. For example, they have a wonderful library of components. They hand annotate it. You cannot do it if you don't have very cheap, highly skilled labor, which we don't have in the US. Like the electrical engineers in the U.S. are incredibly talented and very expensive. You cannot just do that. So how do we do that with a large language model? Like a large language model, if you make it capable enough, we'll be able to replicate some of the mix of ingredients that make something like JLC very successful, in my opinion.

41:21The Generalist:You've taught me a ton in the last 10 minutes that I, after a ton of research, didn't fully understand. So that's really helpful and really interesting. To jump back into the sort of diode story, you have this moment where you sort of realize how little has been invested in understanding, you know, in making circuit boards really have this sort of level of fidelity one would want in the U.S. at least. You and Lenny apply to YC, but with a sort of a different approach. What was the sort of initial version of it?

41:49Davide Asnaghi:The pain that we were feeling back then was I want to catch mistakes before they happen. Right. So like large language models, if you teach them like electronics very carefully in a very calculated way, for example, by representing the schematics as code, representing the layout in a way that like multimodal models understand, there's a lot that you can

42:12The Generalist:catch.

42:13Davide Asnaghi:But and I think that this goes back to the comment that you made about being a copper merchant versus like just an engineer. This was a pain point as an engineer I was feeling. But if you try to sell just this small part of the solution, you are selling vitamins to companies that don't have an express pain point right now. The pain point that this company had was not, how do I marginally make my designs less mistake prone? The pain point they had is, I cannot hire an electrical engineer. I need a board now. And I need it fast. And so in the exploratory space of the technology can do this. I know that this is a problem because I've had it myself.

42:59Davide Asnaghi:Can I sell this? We effectively stumbled into, huh, like if you truly, like in order to train the model that would catch the mistakes, we built a model that would generate boards. And so we effectively use the same infrastructure, but basically say, oh, like we can actually generate the board that you want. Like a lot of the boards that we, like in San Francisco, there's a lot of robotic startups and these robotic startups maybe need some tweaks on some existing off the shelf boards. And this is like relatively easy to do. And so it turns out that people do want this. And, you know, I think that this was like another incredibly underrated like thing about YC.

43:36Davide Asnaghi:Like it was basically like you should follow what people actually say they want. And it has been very humbling, but also like incredibly rewarding because in the process of switching from a solution oriented like mindset to tell me your problem. Let me see if I can fix that. first. We also figured out another beautiful things about companies, which is incentive alignment. Initially, we were thinking about this design platform, right? Just make the design streamlined and more efficient. But if you think about a design company that has no stakes in delivering a product, you're not held accountable to the same standards, right?

44:15Davide Asnaghi:Like if you as a company will suffer the pain of having to rework 150 boards that were designed in roughly the wrong way,

44:25The Generalist:your incentives are incredibly aligned

44:28Davide Asnaghi:to make the best design software that will not do that. And so we live by this tenant. Incentive alignment with our clients has been a wonderful thing.

44:40The Generalist:And you mentioned robotics companies. You have some impressive customers and partners that you guys are working with. To understand the customer journey from sort of design to getting the board, what does that look like through Diode? Like, are they entering your software first and sort of doing some part of it themselves? Maybe sort of just going through that journey would be useful. Of course.

45:04Davide Asnaghi:So this is something that's in flux. And the current process is

45:10The Generalist:we will capture a specification document with you.

45:13Davide Asnaghi:This is partially done on a call. Right now, every one of our customers is incredibly dear to us and we love them very much. And so whatever it takes to make them feel like their spec is being captured correctly, we will do. We also have a version of this where you can effectively explain your specifications to the model. And the model has incredibly high context about parts availability, parts are already being designed and so they can be reused with minimal likelihood of failing. What is in our library? And so we have this, we call it the spec builder. So the first step is as a customer, we will build a specification with you.

45:52Davide Asnaghi:That specification becomes our Bible. That is the source of truth. And then what we do is we have multiple agents that take the specification and do a first approximation of the design. We work with Anthropic to effectively bake some of our design language into the general availability models. We are very invested in large language models becoming better at electronics because our goal is to effectively automate as much of this process as possible. And schematics are on their way to be fully automated. Like I estimate like 90 % of our schematic that we do is like the model can one-shot it now.

46:30The Generalist:90%.

46:31Davide Asnaghi:Yeah. Like schematics are like, they're beautiful because they're very ideal. They're like a logical graph. It's very, very similar to code. Then one of our engineers, we hire incredibly talented engineers because our goal is to actually build boards that work. We'll review the output. And then we have a platform where our customers can confirm that this is what they wanted. Maybe they realized that they said they wanted A, but they really wanted B.

46:55The Generalist:This is the platform that we expose

46:57Davide Asnaghi:at app.diode.computer, for example. I am very proud that this is just a visualization of the code. This is pure, if the yellow lamp touches the code, you will see it in a human-readable way. We generate schematics, which the models have a hard time. Visual reasoning is hard. So we use the code as a hook for the models. we generate schematics for the humans. And then the next step is the layout. So here is where we do a lot of research. Layout is not 90 % automated. I estimate it's closer to like 40 % and 50 % for us internally.

47:32The Generalist:It's like very good at placement, less good at routing.

47:35Davide Asnaghi:We have a few companies that we really look like up to, like a quilter is working on this problem. Like there's an entire company just dedicated to one thing that we do, which is routing. And there's another company called DPCB that we actually have worked with in the past. And both of these companies are wonderful. And I hope that they succeed because our goal is to automate everything end to end. But we do a lot of research in-house. And some of it is non-public, sadly. But our goal is to let large language models become much better at routing. And I think that this will unlock the next layer of productivity for us.

48:10The Generalist:You know, the sort of partnership with Anthropic I thought was super interesting. And it makes sense that, you know, the way that you talk about this, that you want these LLMs to become amazing at these parts of the process because it, you know, just just makes your sort of full stack approach more effective. What does that sort of say about where you think value ultimately accrues in this business? Like, does it become the manufacturing side? Is it just sort of, I don't know, all these things sort of concatenated together and managing that full process? It depends on how AGI-PILD you are.

48:43Davide Asnaghi:Mm-hmm. if you truly believe that the models are going to become radically better at anything that has to do with a computer, there is a reasonable belief that you can hold that the design portion will eventually be just the cost of the tokens that it takes to design this specific part. And the bet that we are making is that first, this is a non-trivial problem to solve. Today, this is not something that just works, right? The models are getting better and we are very happy about this. But this is something that we think will take a couple of years at least before you have full automation on the design.

49:24Davide Asnaghi:The spot that you want to be is a one-stop shop where you can let your model design and then you receive a board. And this is really, I think, where we eventually will want to be.

49:36The Generalist:You said, depending on how AGI-pilled one is, how AGI-pilled are you? How do you build your business around these sort of horizons?

49:46Davide Asnaghi:The beautiful thing is that this is a great business, even at the current stage. The thing I bless my lucky stars for every day is that first, this is what we are good at. We design boards, we bring them up, we make sure they work, and we ship them to customers. And this has a very large amount of value. Even if you're just like 50 % more efficient, Our schematic automation allows us to be 50 % more efficient, which translates to radically faster design timelines. So this works today. And I am very agile-filled. In fact, we are building the company by effectively assuming that a lot of these problems are just getting better.

50:25Davide Asnaghi:And we saw it day by day. Things that were completely out of the question six months ago are now table stakes. all of our library management, which is a very, very sensitive part of electronic development. There's like dedicated person at large companies, the librarian, which will religiously take care of symbols and footprints and data sheets and everything else. This is all done by the model. And maybe you'll see something coming out soon in this domain. But this is working. Like this is something that like lifts hours and hours like from our timelines.

50:59The Generalist:And we can do it on demand.

51:01Davide Asnaghi:So I'm very happy. to see that progress is being made, we are very far from being able to build RF board that is very layout heavy. This still needs a human touch. And we do it for some of our Fortune 100 aerospace customers. And they can compress their timelines from nine months to a month, including the factory delivery. And so this has been just an incredible amount of value

51:27The Generalist:that we get out of this. With these LLMs, with writing or with code, maybe, you know, a better example, it's capable of producing performant code or competent code that sort of hits the bill or meets the brief. But it's not always necessarily the optimal code, right? Like, you know, a truly exceptional software engineer would find a better way to solve a certain problem. I would assume that's the same dynamic for you, but because it's less forgiving, you know, maybe the stakes are higher. How do you sort of deal with that? Like, I imagine there are circuit boards that are viable that a unbelievable electrical engineer would say like, yeah, sure, it's it works.

52:12The Generalist:But, you know, if I was doing it, I would do it in XYZ way.

52:15Davide Asnaghi:The usual telco sign of this behavior is in the routing. The schematics, it's arguably more correct than what... The schematics are... So first of all, they are easier to verify. We have incredibly strong simulation harnesses that give optimization function to the model. This is a core part of what you need to do to validate that a schematic is good. The other part is that schematics are largely, they're a much simpler domain than like pure layout. The ability in which you can compose code is massively easier. Like the tooling around electrical engineering has not evolved so that there are shared libraries of a lot of these things.

52:58Davide Asnaghi:People have tried, but there are a lot of like complicated gotchas that can be solved, but they're fairly hard on the schematic side. No one complains about the validity of the schematics. everybody complains about the aspect, like the visual elements of them. It's a very human thing. Like people are very receptive to how good a schematic looks, even if it's correct. Like you can have a correct schematic that looks bad and people will perceive it as like poor craftsmanship, for example. The way I like to put it is that the universe of boards that we can tackle is always expanding. And today, like, you know, if you start from individual components, the librarian is great.

53:40Davide Asnaghi:And the librarian can also generate reference designs, which are like you can think about as the minimum viable board that you can build around a single component. And now the models are starting to be so good that they can stitch them together and also build the layout. And our bet is that this will just keep expanding to the point that the universe of boards that you can build is so vast that you can just do it all.

54:02The Generalist:Super interesting. I love this idea that often things that look wrong to humans are actually sort of the best version of something. You know, it's like the chess move that makes no sense. And then, you know, 10 moves later actually turns out to be right.

54:17Davide Asnaghi:To be clear, we spend an unbelievable amount of time to make sure that our schematics are very palatable to humans. Like this is a decision that we made at the beginning of the company. It's like, you can take the stance, which is like, basically, don't look at the schematics, just trust the model. And we have a trust but verify stance. our customers will get schematics. They basically get the same output that a human would give them, right? Because there is a human at the end of the day like cleaning up the little edges. And this is what gives them the most value, even as a just archival foreman.

54:49The Generalist:So we do spend a lot of time on that.

54:50Davide Asnaghi:It's not just that we blindly trust the model.

54:53The Generalist:I see. So because humans otherwise would have trouble trusting it, essentially, like there might be a version in the future where... Oh, for sure. Yeah, that's less necessary. But for now in this transition, it's sort of important.

55:06Davide Asnaghi:Yeah, and I think that this is like the difference between building a research project and something that your customer can actually use in production. Like you need to deliver to them a small package that they can open in their EDA tool of choice. Like if they use Altium, they should be able to open in Altium. If they use Kika, they can open in Kika. Cadence, the same thing.

55:23The Generalist:We work very hard to make sure that we are not just a helper tool,

55:27Davide Asnaghi:but a proper productivity suite that we actually can deliver good results on.

55:32The Generalist:You mentioned Kika and Cadence. In the long run, I would assume your goal is to devour that workflow within your platform, right?

55:40Davide Asnaghi:We have a slightly different stance on KeyCAD. Like, I cannot speak about Cadence. Cadence is like a closed software. I'm pretty sure that Cadence cares much less about their PCB software than they care about their EZX software. That's where most of their revenue comes from. But KeyCAD is a wonderful piece of software, which is free and open source. Say that you are an Italian stone maize in Florence or an Italian leather craftsman. The tools are not necessarily what makes the artifact important. The artifact, the end product is really what matters. So our goal is not to devour KeyCAD or Cadence.

56:19Davide Asnaghi:It's to build an automation layer that becomes something that you dispatch work to. You can think about it as we are a layer of automation on top of KeyCAD. And in fact, all of our designs are built in a way where the agent can use KeyCAD as a tool. And we contribute back to mainline KeyCAD. If you go on the repo, we have a lot of patches that we have contributed about manufacturing, fixing some of the problems that we had internally. We are about to contribute another tool. KeyCAD and the tool should be open source. In fact, our compiler is open source. It's called PCB, github-dioding-slash-PCB.

56:53Davide Asnaghi:Fully open source. Our goal is not to gatekeep the compiler. Our goal is to build an electrical engineer that uses the compiler to build boards.

57:01The Generalist:I'm curious where you see sort of like the most natural early adopters here. It sounds like robotics has been one, but yeah, where do you expect Diode to sort of make most headway over the next few years?

57:14Davide Asnaghi:In many ways, like we work with a lot of enterprise customers, which are wonderful and were an important part of the initial set of hypotheses we were testing. It's like, can you deliver value at large enterprise corporations? This is like an important part of building a lasting company. It's like incremental adoption is so important. Every, like a lot of the founders I respect the most have given me this advice. It's like, if you try to build a radically different system by throwing away everything from the old system, you're going to just face an uphill battle. But there's another lesson that you can learn from Stripe.

57:49Davide Asnaghi:like Stripe in the early days had a wonderful product market fit with a lot of developers that did not already have a payment provider. These were folks that effectively were building wonderful things on the internet and needed a new solution that would work for them with them. And you could spend a lot of time like trying to retrofit Stripe into an existing solution or you could go after all these new universal players that just expanded immediately. We want to do both because we think that it is important to make sure that we are good with enterprise customers. But the most fun that I personally have is with some of our wonderful clients in the Series B to Series C range, where they have an incredible appetite for boards.

58:37Davide Asnaghi:They just need to build so many boards. And it's so hard to hire like 100 electrical engineers because the United States just does not have this talent. And the people that are incredibly good electrical engineers are kind of like, they became incredible electrical engineers through a long career. And now they are deservingly resting, investing, or if I am at SpaceX, I'm going to be waiting for the IPO. Like I'm not going to join a grinding startup. Like I've already made it. Like there's a lot of these archetype of people. And these are incredibly smart people that you want to learn everything from, but maybe they don't want to be like going back to square zero, like build a hundred board from scratch.

59:14Davide Asnaghi:And so this is where we come in. These are the people that we have a ton of fun working with. And we will do whatever it takes to design those 100 boards with them. And then we hope to become the best partner to bring them from new product introduction to actual large-scale manufacturing. We have a lot of partnership with people that are set up for the world of, here's one design that's done. Let's build 100 ,000.

59:38The Generalist:And so even before we have the ability to do it, this is very largely a prototype line.

59:43Davide Asnaghi:This is not a line that will scale to millions of units. But this is something that we can do for our customers. And we want to be that type of provider. We want to be the new generation like Stripe, but for electronics and for hardware.

59:56The Generalist:I think every great company ends up having at least one near-death moment. Have you had to go through one of those yet with Diode?

1:00:04Davide Asnaghi:I am very fortunate that, like, I think that the near-death moment was in the very first few weeks where because Lenny and I were just like, we are very deeply like the engineering mindset is very deeply ingrained in both of us. And as an engineer, a large portion of what you are trained to do is someone will give you a problem statement from A to B. And your goal is to get from A to B in the most efficient and optimal way possible. Like what I have found, like building a company is like is you are given A and you're asking yourself, what is B? Not even where it is. It's like, what type of shape will I have in B?

1:00:48Davide Asnaghi:And so like figuring out a business model that allows us to go where I want to go, which is I want to see the United States be as nimble as it is in software. Like the United States has such wonderful software talent. This is probably the thing I'm most in love with about this country. It's just like you come here and you walk around Berkeley and you see people that are so talented in the software domain. And they just find hardware unapproachable enough that they find it hard to actually go and apply those talents into that domain. I need that to not be the case. I want the US to become as nimble as hardware as it is in software.

1:01:30Davide Asnaghi:This is like the third tier description of Diode. Like if Diode wins, what we will see is much larger portion of talented and smart people being able to whip up electronics the same way that they can spin up a SaaS server. This is what we want to achieve.

1:01:46The Generalist:So in that sense, you know, every company is unique. So making comparisons isn't always helpful. But for the sake of sort of clarity, it almost sounds like the ultimate goal is an American Foxconn for the AI era. How close am I there?

1:02:04Davide Asnaghi:I think that you're incredibly close. I think it's also very flattering of being compared with Foxconn.

1:02:12The Generalist:These are the aspirations, right?

1:02:14Davide Asnaghi:That's right. There is a lot of work that needs to be done. And the work that we are focusing on today is wonderful design and design for manufacturing platform that holds these two ends of the problem together. there.

1:02:29The Generalist:I'm curious about you and the way you run the company as a CEO. What is the sort of culture of diode that you care about? And what are the things that maybe you've chosen to do differently than some of the other places you've worked?

1:02:44Davide Asnaghi:I think that there's actually a lot of like other places that I've worked. I don't know that I have anything reactionary. There's a lot of like picking the parts that I love the most about different places that I've worked. I think that I am obsessive about the talent that we bring in. I spend a lot of time here. And when I spend time here, I love doing it with people that are engaging. And I respect incredibly highly. And I are fun to be around. And I'm very, very proud of the entire team. So I think that we have been very deliberate about hiring in a way that like the bar is very high and we have so much work and it would be so easy to just you know give in and be like okay like we might just hire like 20 people because there is enough like work but I think one of the things that we have been doing slightly differently is we have been focused on where is the pain like where is the pain coming from where is the thing that is not scaling like it used to be schematics and now schematic is gone And now we see layout as the next big hill to climb.

1:03:55Davide Asnaghi:And that signal has been very strong. But we are hiring. And in fact, I am hiring very aggressively. So if anybody listening is incredibly excited about electronics, and you are either a very talented electrical engineer or a super smart software engineer that loves to dip their toes in the hardware pool, please, we'll fly you to New York. Come see the factory. Like this is like truly, I think like a hallmark of building a good company is like how excited are you to see your co-workers? And this is something that I have like been blessed with, like being very excited about it many, many times in the past.

1:04:32Davide Asnaghi:And this is the number one thing that we want to hold true. I think that the differentiating factor is in like how high we keep the bar for hiring.

1:04:40The Generalist:Well, hopefully we book you a few factory tours from this. Also, leaving New York, big perk. There you go. Yeah, absolutely. Where do you find you come up with your best ideas? Like, are you still, I don't know, fiddling around with electronics in your spare time to understand these things? Is it on a walk, you know, sports, whatever it might be?

1:05:02Davide Asnaghi:I would be truly unhappy individuals if I stopped building boards. There was a mandate of building at least one board per week. Like, I need to be the least productive electrical engineer at the company with one board a week.

1:05:18The Generalist:Okay, there you go.

1:05:20Davide Asnaghi:I am the low bar. I am the worst E at Diode and I still need to be aboard. And I think that this is starting to be harder because we have incredibly high-stake customers and we want to deliver a wonderful experience. So what I'm doing today is I'm building a lot of the hardware that goes into the factory. A lot of the problems that we find is that the machines that you see are mostly meant for manual operation and we need to figure out a way to truly automate them from figuring out the right data model, figuring out which piece of hardware can talk to what. Even the conveyor belts, they are not particularly smart.

1:05:54Davide Asnaghi:How do you make them smarter? How do you stop them, take pictures, stuff like this? So I still build a lot of that hardware. And I think that the best ideas actually come from talking to any. It is truly underappreciated how much talking with somebody much smarter than you kind of unlocks new modes of thinking. I think that I am more prolific in the set of ideas that I generate. And Lenny is much better at discriminating which ideas are worth pursuing. And so this dynamic has helped me so much because he will basically say, like, okay, you have 20 ideas. The rate at which you're generating it is probably going to kill the company.

1:06:34Let's focus on these three.

1:06:36Davide Asnaghi:And I'm like, no, no, but I really like five. And he's like, why do you like five? And then we discussed, like 90 % of the time he's like, you were right, five is probably something that we can kick to later. And 10 % of the time, I'm like, oh, actually, we get to do five. That's fantastic. And I think that that has been, like, I bless my lucky stars that Lenny is my co-founder because it's a wonderful person to have on this journey.

1:06:59The Generalist:Amazing. Well, Davide, I've enjoyed this so much. And I always love to end with a few sort of more thought experiment questions for you. If you had the chance to run any experiment with no operational constraints and unlimited resources, What's an experiment you'd like to run?

1:07:17Davide Asnaghi:Is this in the realm of just like Diode as a company? Because there's a lot of like larger experiments that I think like should be.

1:07:23The Generalist:No, no, it should be as large as you'd like.

1:07:26Davide Asnaghi:Dog Patch in San Francisco needs to be the new, like this is the Brooklyn Navy Yard. It's like a part of New York that was like developed like in the, I think 50s. And now it's like an industrial space that just overlooks the city. I need Dog Patch to become like the Brooklyn Navy Yard of San Francisco. I think that largely there's a lot of companies that we know in the industrial sector that are there. And I think that like Astronus is there. The American Industrial Center is there. One experiment I would love to run is you take the entire surface and you make it like an industrial city. I think that that would be incredible for like bringing Silicon Valley closer to what Shenzhen is.

1:08:06Davide Asnaghi:And I think it will happen. And hopefully we are a part of making that happen eventually.

1:08:10The Generalist:Okay, final question for you. had the chance to assign a book to everyone on earth to read and know that they would understand it, what would you want to assign?

1:08:20Davide Asnaghi:There's really two. And I think that they should be done in sequence. So if I have to pick absolutely one, it's going to be Chip Wars.

1:08:27The Generalist:Yeah, it's so good. It's just no flaws.

1:08:31Davide Asnaghi:Before starting the company, I spent the year prior reading a ton of both domain expertise books, like Chip Wars and science fiction. I think it's tremendous how much science fiction actually captures what the ethos of progress was like felt like like 1970s science fiction like what did people 50 years ago like think but i do think chip wars is a wonderful primer to understand truly and deeply how much american government like the american government has had this outsized influence in developing silicon and how and why like ram production was outsourced to Singapore and Vietnam and like Taiwan and like how it happened and how like Japan was like this incredibly strong powerhouse.

1:09:19Davide Asnaghi:And then like the balances of power that have shifted over the last like 60 years. I think that if you do not understand the geopolitical angle, it's very, very hard to understand reindustrialization. Like it's very easy to devolve into like us versus them mentality, like for many reasons. But I do think that if you do understand the incentive structure that operate at a nation state, you can understand why even if you are a full pacifist and you don't care about rivalry between geopolitical rivals, having a strong industrial base is just important for your long-term prosperity. And this is something that I think Europe has not done particularly well.

1:10:02Davide Asnaghi:And I would love to see more. Like, I think that Europe has such, like, all these machines are European, by the way. Like, it's wild. Yes, it's wild how many, like, high-end industrial machinery is still made in Europe. But there's this weird middle ground where it's like, I want to transition to a services economy, but then I have, like, these, like, small and medium businesses that do manufacturing.

1:10:27The Generalist:I think that this is not a contrarian opinion anymore. But, like,

1:10:30Davide Asnaghi:understanding the geopolitics is just very, very important to understand how the next 50 years will play out. And why China has been like largely so successful, which is, I think, something that it's worth learning and admiring the good parts of and recognizing where like there are bad parts that you can avoid.

1:10:47The Generalist:Okay. So chip war is a recommendation one. You said there's two. What's the second one?

1:10:52Davide Asnaghi:Breakneck is just like in order to like, and this goes again, like Like demonizing like a country like China is bad for everybody. Like you should yearn to understand it. And then once you understand it, you can have a nuanced opinion. And there's things that you can denounce and there's things that you can admire. But you truly need to understand like the idea of an engineering culture and an engineering nation is very profound and very powerful. And like very scary in some ways, like social engineering, for example. But like Breakneck is, I think, like a wonderful book to begin and understand that and like be able to capture the good parts without like throwing the baby with the bathwater type of thing.

1:11:37The Generalist:Well, those are two great recommendations that I feel like capture a lot of the spirit of this conversation. So thank you again, Davide. I really enjoyed it.

1:11:45Davide Asnaghi:I enjoyed it too. And thank you for having me. This has been wonderful.

1:11:49The Generalist:That's it. Thank you for listening to this episode of The Generalist Podcast. Please subscribe on Apple Podcasts, Spotify, or your preferred podcast app. Ratings and reviews help others discover these discussions, so if you enjoyed the conversation, I'd be grateful if you could take a moment to leave one. For all past episodes and more, visit us at thegeneralist.substack.com. See you next time as we continue to explore the future.

1:12:23Thank you.