David Stålmarck

This transcript was generated automatically (YouTube auto-captions) and then edited for readability: names and obvious mis-hearings corrected, filler words removed, speaker labels added where the speaker is clear. Passages that could not be understood are marked [unclear]; short omitted passages are marked [...]. Timestamps are approximate. The recording is the authoritative version.

David Stålmarck 00:00:01

We want to make it as easy to build hardware as it is to build software.

It's almost like Nvidia and Lego had a baby.

David Stålmarck

So we did get quite a lot of eyes on us and many visits in our DMs. We were told by the people here, like, take no [unclear]. Don't speak to any VC.

Host

A crazy thought experiment to just ghost the VCs in your DMs and be like, "We're confident enough."

David Stålmarck 00:00:24

Decentralizing who owns the hardware when physical AI becomes an actual thing.

Host 00:00:31

Then hello and welcome to a new episode of Campus To Capital. And today it is an episode that, yeah, I think we're both genuinely excited about. We're in the Atech office now, or like the [unclear] office if I'm like 100% correct. And actually I've been there yesterday as well for your, yeah, I would call it launch party. It's also a story like how we got to know David because we've been at a hackathon last year and met you there and then somehow caught up and we thought it would be exciting times to record an episode. So yeah, short hello and welcome David Stålmarck, founder and CTO of Atech. It's so nice to talk to you today.

David Stålmarck

Thank you very much. I'm super excited as well.

Host 00:01:19

Yeah. Hi from me as well. And I'm also very excited because I mean we cover a lot of topics, a lot of guests that engage in, let's call it, the technical revolution that we're facing right now. But actually hardware is something that I'm personally quite unfamiliar with. So I'm excited to learn some new things today. So let's get started. And talking about hardware, what are the biggest bottlenecks that make hardware development still difficult to access today?

David Stålmarck

Good question. There are a few things here that make it difficult to work with hardware. The one that we are mostly focused on is zero to one, the first steps, and then it's just that it requires a lot of knowledge on how to do it and you kind of have to know this knowledge. You also have to buy all of the parts, you have to get them, order them online, it costs money, there's friction there, and then there's also friction that there's a risk of accidentally destroying it. You also need a big upfront cost with like having a soldering iron, a bunch of spare parts, a 3D printer, etc., other gear, and so only people that are really committed will start with this. And with really committed, I mean also like studying in an electrical engineering degree or just because they think it's really fun. So that's one thing, like

00:02:49

it just takes a lot of time and resources. And then the fact that we have LLM-assisted coding, or AI can do so much for us when it comes to software, with hardware it can't really assist that much unless it has all of the context, like every single little detail on this board, the full schematics, and for it to be useful, to be able to get the like superpowers that we got from LLM-assisted coding to hardware, we need to be able to pass that context as well somehow. And that is like getting started with hardware. Then we have the full supply chain side with being actually able to produce a hardware solution. And right, hardware iterations is a completely different story as well. In software, you can literally just rewrite the code instantly, test it. In hardware, if you're designing your own printed circuit boards, you have to design them, you have to order them and now, like, more or less the only place where you can order proper ones is from China. There are like some super fast ones that can deliver it to you within a week, like that's amazing, however that's still a week for one iteration, which is like two and a half minutes when you're in software. So

00:04:19

the iteration speed is also a big limitation.

Host

Okay. So, it's very capex and time intensive, right?

David Stålmarck 00:04:26

Yes.

00:04:26

That was a good summary. Yeah.

Host 00:04:28

So I mean many of our listeners might not know what you're doing because you were working hidden I would say for a couple of months and now launched very recently. So you described it yesterday. I think someone from Lovable described it as the Lovable for hardware, right? So could you maybe, I see some hardware in the background as well, could you maybe show our listeners essentially what your products look like and what you're doing?

David Stålmarck 00:04:57

Would love to.

00:05:00

No, but so we want to make it as easy to build hardware as it is to build software. For that we have taken inspiration from software and how we have different kinds of abstraction layers on top of each other and libraries where we can just import the library and we can use all of the functions from there, but in our case we made a bunch of modules that would represent these libraries. So this is a distance sensor that measures the distance. And instead of having to like figure out, okay, how do I connect this one, how do I write code to it, you just click it on and it works out of the box and you're able to prompt it in our platform how to use this. So now you have a distance sensor, you have Wi-Fi that you can send this information to. So suddenly I can now from my computer know, oh, what is the distance here. I have a button, I have a screen, we have here some motors, we have a bunch of different modules and you can quite quickly do like a Wi-Fi controlled boat where, yeah, you don't need any kind of electrical experience.

Host

And I mean one of your co-founders showed that it's not like that, it's also possible in real life, right? Like in real measurements. So Gustaf,

00:06:31

he, I saw the video that he, yeah, sailed from Sweden to Denmark with an Atech controlled, what do you call it? The

David Stålmarck 00:06:41

they call it autopilot.

Host 00:06:42

Autopilot. Okay.

David Stålmarck

And, or I got to learn there as well on the boats, but the autopilot, the current forms of autopilot, or the one that they have, like an actual autopilot, or how do you say it, cost $2,000. So it's a private [unclear] thing and what it does, you put it into a certain direction using a compass and then it will just move the rudder to follow that direction. So it's, how do you say, it's not that you press a location on Google Maps and it goes there. It's a direction. But it's basically the linear actuator that will move the rudder. And this he rebuilt then using an LTE board and one of our more powerful motors that we have.

00:07:31

It's almost like Nvidia and Lego had a baby.

00:07:33

Yeah,

David Stålmarck 00:07:34

that's what we're trying, to make it as easy and as intuitive as Lego to

00:07:41

build the hardware

Host 00:07:42

which is very cool. But I mean you already mentioned Lovable and now when we quickly go back to software, I mean up until recently you didn't have tools like Lovable and it was very difficult and very skill intensive to build software. I mean advanced software still requires a lot of skill today and also we talked to another co-founder in Copenhagen yesterday who's focused on software. We also talked about that it's a very capex intensive endeavor. So thinking about these constraints that also were in the software field, why do you think there was so much emphasis on developing software and changing how we think about software compared to where we haven't seen such a big push for disruption in the hardware landscape?

David Stålmarck 00:08:30

But it's, understand that people talked about that software used to be very capital intensive and there has been a big push to make it less?

Host 00:08:38

Well no, I mean it's still, it's capital intensive today, but rather that like we see a lot of new advancements in software and I think like hardware hasn't been talked about that much compared to software. So in your eyes as a hardware founder, why is that the case?

David Stålmarck 00:08:53

Well, it's funny here because I come from a software background and I've joined this team basically as the software guy. My two co-founders had a hardware company for two years. I've been, yeah, working at different kinds of tech startups and big tech companies and mostly done a lot of coding competitions. So I come from the software side and what I handle here is our software. So it does take a lot of time and it is also like the vast majority of things that makes it magical with hardware is not the hardware itself, like the precision of these motors or the distance sensors or the screen. That's not what is cool, to just get the measurement and send that. It does take a lot of time if you have never done it before, but that is not what the cool physical AI thing, the physical AI thing that people talk about, is the software integrations to it, that you can just prompt in our platform and suddenly get this data to your platform. You can quickly integrate it with whatever system you want and suddenly you can facilitate, like, every time a door is opened or whatever you want to imagine. So I just think there's a lot more things

00:10:23

to do in software, and for hardware companies as well, where like most robotics companies are software companies that, like, we help you get data for training [unclear] for robotics etc., or we help with scheduling how robots work in a factory. Most of the things are software still. So therefore I think it's more capital intensive and more focus on that. It's a better business and, yeah, on the better business side, it's also you can just sell it instantly. There's not this supply chain question. You don't have to go to Shenzhen [unclear] once a month and check up on the relationships there. But it's really coming and we have seen multiple, I guess, competitors to us. Funny enough they only do software as well. They do software that tells you how to build it with the current hardware stack, but what we add to this is just to make the assembling part easier as well. So

00:11:34

so

Host 00:11:36

essentially the users could then go on your platform, right, buy a set of, I don't know, 8 to 12 components, then they get, that's going to be delivered at home and then they can prompt, okay, I want to build an alarm clock, and then it will give the perfect, you know, how to build an alarm clock with your components, and that's how it works.

David Stålmarck

Yeah, actually, yeah, splitting our product, yes.

00:12:03

I'm gonna answer, I got the long track [unclear]. No, but with software in general, that's the main thing, that's what makes hardware cool, the software behind it. We don't care about how high the precision of the distance sensor is or how high the precision of the motor is.

00:12:27

We do care that we are able to assemble and, like, the reliability and quality of it. Yep. But apart from that, what makes software cool, and what this like physical AI robotics hype that we have currently is, is the software enablers. The new models that are focused on robotics, and people are talking about like we are soon going to have some kind of ChatGPT moment for robotics.

Host 00:12:55

That's going to be exciting.

David Stålmarck 00:12:56

And all of that is just software, right?

Host

Stepping out of your shoes as the co-founder for one second and maybe viewing your product through the lens of a potential user, because you just mentioned the alarm clock, like what's your wildest dream and imagination, like what's the coolest thing that you are thinking of that you want to build with Atech one day?

David Stålmarck 00:13:16

I really want to get into robotics. I think that's super interesting and I think it's really important to understand how that works. And yes, for multiple reasons and for multiple, like, bigger humanitarian goals, etc., like

00:13:41

anything from cleaning the oceans to building houses in it,

00:13:47

it comes down to

00:13:49

robotics, long-term solution.

Host

I mean, I think at the last Nvidia investor day Jensen Huang, I mean obviously at those types of events there's a lot of CEO fluff which is not super exciting, but he also talked about that he has the vision to eventually outsource factories to the moon and then have robots working on the moon, and then eventually, like this is very far-fetched now, but he was talking about eventually that we get to be able to travel with light speed and then we have robotic factories on the moon that do all the labor intensive work for us. I was like, I mean then we are beyond Star Wars if we get there, like that would be so cool.

David Stålmarck 00:14:28

Yeah. I agree 100%. But also it would help us a lot more in this utopic future that people envisioned, where humans can just do what we want to do and we have the universal basic income basically in that scenario, at least how I'm envisioning it, that we are not doing our dishes, we are not folding our clothes [unclear], and all of it is simple. Also for like industrial use, these handmade woven things, we're not doing that unless it's for fun. Also speaking now, like, in the future, I don't know how far in the future, but we have to be able to control the physical world, and that's software, and I've used [unclear] robotics a lot more as a software problem than a hardware problem.

00:15:27

Yeah, of course, like these crazy cool different kinds of hardware solutions, like Boston Dynamics has with like 360 full degree rotation for every degree of freedom, that's super cool hardware solutions, but yeah, entering that space and making it easy to, I'm not a big believer in humanoids for that matter,

00:15:53

big believer in robotics but not necessarily humanoids,

00:15:56

to be able to build different kinds of custom solutions to whatever problem. If it is some kind of home automation or if it is more a business use case, you want to automate your farm or your pharmaceutical lab. We want to be able to basically build internal tools for factories. So that's I guess what I'm most excited to try to do.

Host 00:16:19

It's a cool vision. So I really hope that it's the utopian future that we will get to experience one day. But then going back into the present and where we are right now, if someone now goes about using Atech, have you already somewhat understood how much faster it allows them to prototype their hardware idea and how it accelerates their process? Like is there already some number, how much people can speed up their timeline potentially?

David Stålmarck 00:16:49

It's a hard question to answer with like an

00:16:53

it's very [unclear] for sure. So if you want to do something that fits perfectly for using, depending [unclear] on what your present is that you're comparing to. So I guess for most people it would be an Arduino starter kit or something.

Host 00:17:11

No.

David Stålmarck 00:17:11

And assuming you already have all of the components. This is also like, you can't just pip install or download something from the internet. You have to actually have it. Well, yeah, you can try it yourself. The assembling part is like seconds or minutes. And then if you want to iterate and test something new, let's say you're building a boat like this and say, "Oh, wait. You know what? I want to add another distance sensor on the sides so it can ping if it's too close to something." That takes more or less a minute.

00:17:51

Yeah. And it would probably take, I know, with the Arduino 20, 30 minutes. If you don't have the parts and have to develop it yourself and have to make a new printed circuit board, well then it's weeks, because you have to order that from China with a new design. So depending on what your current status quo is, like very huge speed-ups in hardware iteration

Host

and probably even much higher potential, right? So I mean you already talked about that, yeah, you basically are reliant on China as they're like the only feasible producer for your tech boards for example, and that you send out one or two of your colleagues there to negotiate. So I was wondering, because we also talked about these iteration cycles etc., how much, because I talked about this with people yesterday actually, we were wondering how much is it a logistics problem for you and how much is it like actually developing the problem, like what are you worried about the most or what are the highest risks for you in that regard?

David Stålmarck

A very good question. Something we think about a lot too. But I think two are the most pressing, or for any early startup that they're building something

00:19:21

cool that fits for a lot of things but doesn't fit perfectly for something yet: finding product market fit. It could be like something super niche in one industry where they really want a modular design. For instance, biologists, researchers that are doing a bunch of experiments, to just quickly iterate around, or doing autopilots for sending [unclear] boats. Who knows? Something specific, but finding a proper product market fit where we can sell a lot and be profitable, and not being dependent on investor money, is one pressing thing for us. And then with the supply chain, to be honest we don't know how difficult it's going to be, and right now, because it's only been small quantities, it has worked fine, but let's say we do find product market fit and we do want to supply like 100,000 boards and some specific modules to it. We are doing everything modular, also how we are designing it internally, doing everything modular to, yeah, have this flexibility and iterate faster for us. What if they just run out of our connectors that we use to connect them? Oops. Then we just have to [unclear] the whole thing or we have to find someone else

00:20:51

that can produce them. But if no one else can, then we, oops, are at the [unclear]. So I guess finding different kinds of technical solutions so we can quickly protect ourselves from those kinds of supply chain risks, or yes, to figure out how we are having that, or securing very large quantities in advance, but then we need upfront money.

Host 00:21:17

Yeah.

David Stålmarck 00:21:18

And finding good product market fit and not getting lost in the space of just trying too many different things. Well, I guess like we had time. We had

00:21:30

to do that. That's I guess within a year or so.

Host

Okay.

David Stålmarck 00:21:33

At least.

Host 00:21:34

So, that was actually also something I was about to ask. I don't know if you're allowed to answer, like these contractual agreements with the producers in China, like for how long you have those contractual agreements or if it's like, I don't know, one-time orders, like in very simple terms, how does it work? I'm just trying to, you know, get below the

David Stålmarck

Like right now they are so streamlined that you can design a PCB and just more or less press order.

Host 00:22:09

Okay.

David Stålmarck 00:22:09

And they will make it for you, so like there are different kinds of suppliers

00:22:17

and it's more, I guess this comes in hand when we are ordering larger quantities. Now we're ordering in the hundreds when we order stuff. It's still like

Host 00:22:28

Yeah.

David Stålmarck 00:22:28

relatively small.

Host 00:22:30

But that's so cool that you can do it on that little scale. Because that involves less risk for you as well, right?

David Stålmarck 00:22:36

100%. Yeah. And like what we have, we have just ordered like 10.

Host 00:22:40

Okay.

David Stålmarck 00:22:41

And then check that everything works. This is back to what we were talking about with hardware iterations.

00:22:45

Yeah. Prototyping, this zero to one, like crazy speed-ups with Atech, and then when it comes to building your proper product, whatever you want to build, we are working on the form factor, that you can have it in different form factors. Today, like 5 months in, it looks like this and it is like building blocks. You cannot build a smartwatch today or something, but we want you to be able to do that in the future. But when you're building something more custom and you have to design your own boards, like how it is for us when we're designing this, yeah, then the [unclear] comes in again and you have to first order them in small quantities and then when everything works, woo, let's go.

00:23:34

Let's scale.

Host

Yeah. Nice. Moving on from maybe the relationships in China, going to more our own home turf and relationships in Europe. Last week, I think it was last week, I mean that was a very impressive LinkedIn post where you highlighted your supporters and backers and one of them is Lovable, which we've talked about already today. And how did that relationship with Lovable initially come about and also what made them believe in your vision so early?

David Stålmarck 00:24:05

Yeah. So to the first question, like how did we get to know them more or less? Since I grew up in Stockholm and I've been very active in the competitive mathematics and competitive coding community, which many of the people working at Lovable have been, and also Anton Osika has also been, how to say, recruiting from these coding competitions and math competitions for a long time. So I remember texting with him back pre-COVID or during that time.

Host 00:24:42

Okay.

David Stålmarck

From these coding competitions, and then had like small contacts, [unclear] hackathons. And then it was one of my friends that is running some kind of, how to say, AI safety lobby organization, talking a lot about universal basic income, funny enough. But he works there and so I was a [unclear] as well and, like, hey, can you come in on the meeting? So I sit there in this meeting a little bit like this and we're talking about completely different things and I'm wearing a Build Hardware shirt. It was like the first Build Hardware shirt that we had and we had just been at an event earlier and he just walks in and points at me, reads what it says, Build Hardware, and yeah, then asked what we are doing. So like pure luck, being at their office

Host 00:25:45

and wearing the right clothes,

00:25:46

cool t-shirt.

David Stålmarck 00:25:48

And then he asked about it and he's really about enabling everyone to be able to create things that they want and that you shouldn't have technical barriers and friction for the solutions that you want to create or for the problems that you want to solve. So that's a lot what Lovable is about.

Host

Yeah, I was about to say, in that sense your philosophy matches quite well with theirs for sure.

David Stålmarck 00:26:14

Yes. So he thought it was super cool and awesome and like, okay, call your co-founders now. Call your co-founders now. We will speak. And then, yeah, we met up like the day after again and then we met up the week after that and sorted everything out.

Julius Ambrosius 00:26:32

Super interesting. I think the biggest difference probably between you and Lovable is that Lovable tries to make software accessible for the 99% that don't know how to, and you're making it accessible for the 99.99% that don't know how to do hardware, right? I mean talking about investors, Konsti and I yesterday had a podcast recording with also a pre-seed startup that had a pre-seed raised by prominent investors as well and we then talked with the founder about how important this signaling is early on, like who leads your pre-seed, like who's investing. How was that experience for you? Because you also got investments from scouts from Sequoia and a16z, right?

Host 00:27:22

Which is pretty cool by the way. Congratulations. That's amazing.

00:27:26

We talk about it like it's the most normal thing, but pretty stellar.

David Stålmarck 00:27:32

No, but I think we were quite good at building in public. We would post like almost every day. So we did get quite a lot of eyes on us and many visits in our DMs. We were told by the people here, like, take no [unclear], don't speak to any VC, so we didn't really and we postponed all of that, but then suddenly we accidentally spoke to one and then things got very fast. But so for us then, so that was a Thursday, the day after, Friday, yeah, we happened to meet Anton Osika, took a selfie with him. That didn't help a lot. A bunch more VCs started contacting us. But yeah, then there was one VC that had contacted us from before a bunch of times. We just hadn't answered yet. But someone here recommended speaking to them. We called them at like 1:30 a.m., like in the middle of the night, and they just jumped on a call and were like, "Hey yo, what's up? Yeah, we've been wanting to speak to you. We think what you're doing is super cool. Let's have a meeting tomorrow at 10:00." And we were like, "Yeah, sure." At the same time, like we

00:28:48

started speaking a little bit with this. We didn't really know where to go. We got a term sheet very early, which gave us confidence, I think.

Host 00:28:58

Mhm.

David Stålmarck 00:28:58

And then he brings on, so we speak at like 1:30 in the night, at 10:00 a.m. we have another meeting with him and he brings on Felix H. [unclear], yeah, to the meeting, and like that did impress us a lot, like, wow, he's working hard to get this. This fund called Emblem [unclear], which is one of our lead investors. So then we went with them and I guess it was kind of them that got it to us and I guess they did a very good job in convincing us we should go with them.

00:29:34

So we got the [unclear], and Felix thought the product was super cool as well.

Host

What was the reason for the initial advice not to speak to VCs? Because like it may not allow you to [unclear] as freely as you would like to in the start?

David Stålmarck

Or they have a lot of opinions and it takes a lot of time and they are trying to get data points on your startup. This is what we were told, and like the more time you give them to think about it, the more the risk of them being hesitant is. The more pressure you put on them

00:30:14

the faster they would make a decision, and it's also a little like a mind game in the sense, once someone wants you, or some people want you, everyone wants you, it really felt like, and how people described it for us. And if someone doesn't want you and says it out publicly, that discourages others as well. A big FOMO game. This is how we had it described to us. And then just in the beginning, just build, build, show it to people that want to use your product, but just build, don't speak to any businesses. You are going to be able to speak to them anyways. It's not like you lose your chance if you don't take this meeting at 10 on a Thursday.

00:30:56

They're only going to want you even more.

Host

Exactly. A crazy thought experiment to just ghost the VCs in your DMs and be like, "We're confident enough."

David Stålmarck 00:31:04

We did, like, politely write to them like, "Sorry, we're trying to build right now. Happy to speak soon in the future." Like, that's why

00:31:17

it is douchy to. No, of course. No, no, no. But I think

David Stålmarck 00:31:21

I try to a lot as well. No, just like forgetting to answer, but when we did not forget to answer.

Host 00:31:31

So interesting. I mean unfortunately we have to slowly wrap up. I think we could, yeah, keep talking for at least one hour.

00:31:39

Got to respect the founder lifestyle.

Host 00:31:41

Exactly. So maybe to round up the very interesting and insightful, like for us insightful, conversation we had today. We have a lot of listeners that want to try out a venture themselves. So as now you being at the very early stage, being quite young as well, what advice would you give them if they want to start out, maybe one advice, if you have to narrow it down to one advice, what would you say?

David Stålmarck

Building in public. That worked very well for us and like we didn't have to think about getting momentum because it just happened by building in public. And then I think, for me personally, before I found my co-founders, I was basically [unclear]. I wanted to start a startup for almost one and a half years and I would just attend every single hackathon I could find, and alone, to meet new people.

Host 00:32:47

Hackathon maxing.

David Stålmarck 00:32:48

Hackathon maxing and building in public.

Host 00:32:51

Good advice. And maybe one last very quick thing and you can answer this as briefly as possible. If Atech becomes successful, what is in your eyes the biggest ideal impact that you guys will have in maybe, let's say, 5 years in the future?

David Stålmarck 00:33:09

Decentralizing who owns the hardware when physical AI becomes an actual thing.

Host 00:33:16

Awesome.

00:33:18

Cool. Thank you so much. Appreciate the conversation. We'll let you go to your next meeting now. Great speaking to you.

David Stålmarck 00:33:24

Yes. Thank you. Thank you guys.