 The gigatron, but I will not really be focusing on the workings of it. By the way, this is a gigatron computer. So I will not go into the very nitty gritty details. There's existing talks that I did before that you can find on the internet. I will briefly discuss what it is, and then I will go into the process of making a kit to help people who might also be interested in making and selling then their own electronics. Do it yourself, kid. So it's about the lessons that we learned in the gigatron project. By the way, is there anybody who here has heard of the gigatron? Quite a few, not everybody who has a gigatron to build it. Very good, very good. So first, a little bit about the history, how it came about, a short introduction. And then I'll go into the whole kit making process. Actually, there's plenty of kits out there. And am I still audible? Yeah, they are fun to make. They're fun to use. And for some people, that's where it ends. I mean, you can have enough fun just getting such a kit, building it, and tinkering with it. That's perfectly fine. If that's satisfying enough for you, no problem. But this talk will be about creating a new kit that didn't exist before. It all started a few years ago in the end of 2016, not by me, but by myself and Kervink, a friend of mine, who sadly passed away two years ago. And he bought some TTL ICs, chips, that you can see here, a soldering iron, an oscilloscope, and a big book about discrete logic. And he thought, let's do an interesting project. And I joined the project six months later, which I will explain further on. Marcel's first objective was really to, as a hacker, to learn and have fun with electronics. And he thought, well, let's build a CPU. So he had this bunch of 7,400 series TTL chips, which are chips that are really basic building blocks, like and or nor stuff. And he said, well, let's build a CPU. And that's something that a lot of people have done already on the internet. There's a CPU web ring, where you can go to all the websites of the people making such CPUs. So the goal was not to be useful or make really something that other people could use. It was just a learning project. And the inspiration throughout the project was a very big inspiration was Ben Eater. Ben Eater has this wonderful series on YouTube about CPU design. And he uses breadboarding. And he builds a computer from the tiny beginnings to a whole 8-bit computer. He's created this breadboard computer. And that was an inspiration. Also, it shows that you can actually do this. And I will go into this in a little bit. Also, the Quark 85, which is just an 80-tiny 85 that bitbanks a video signal out of it, a VGA signal, was also an inspiration and was a key factor in defining what the gigatron was supposed to become. It started with this system in early 2019. And this is a 4-bit CPU. So this was Marcel's first attempt at building his own CPU. And there's a big chip on there. That's an ALU, an algorithmic logic unit, and a few other components. But actually, it filled kind of miserably. And after January 19 came this Quark 85. And this inspired Marcel to think, hmm, can I also do some VGA bitbanging? That would be mighty interesting. Because the idea at first was to make a simple CPU that could flash on lights or a tic-tac-toe. But this bitbanging was interesting. So he built the thing on the top right, which is a system that, with simple logic, takes an image out of ROM and puts it out on VGA. So it's just like the Quark 85, but not using an 80-tiny, but using very simple electronics. And that worked. So that was the first step. And then out of that, he proceeded to not use ROM, but use RAM. Because in the end, you don't want to have a static image from ROM. You want to be able to change the image. So on the bottom right, there were RAM chips, which he had to load using a microprocessor, an external board that you can see on the breadboard as well. And then after a lot of work, in June 2017, there was the prototype of the gigatron. So this was the first working breadboard version of the gigatron. And that was the main proof of concept. What's interesting is that you can actually have a look at this project, because it's in the home computer museum of Bart van Agar. And so if you go to Helmond, you can go around this fabulous computer museum, play with all the computers, and also have a look at this breadboard system that's over there. I'm quite proud of that it's an actual museum. Now, the gigatron rules that were made during the process were we want to have a CPU that has no complex logic at all. It should be all really simple stuff, like stuff that you could buy in the 80s or 70s even. So no ALU, this big hunking chip that does all kinds of arithmetic. No, it's not in the gigatron. That's done with simpler components. It should be a single board computer with about 34 chip count. And that was inspired by Steve Wozniak. Steve Wozniak, his first job at Atari, was to build a game console arcade. And he was asked by Steve Jobs to reduce the chip count. It was like 100. And he was able to reduce it to, I don't know, 30 something. In the end product, it was like 40 chips. And it was really a big achievement of Steve Wozniak to make such an interesting system with such a small amount of chips. But we still wanted to do something useful and have a nice look and build to people. But if you want to build something that's interesting, as I said, you can just buy an existing project. And if you do it yourself, it will be nice if you build something that stands out, that you make something unique. So that was also something that was in this kit. It needs to have something that others don't have. So find a niche. And by having this system, which is actually a CPU combined video card, makes it a unique system. A lesson in the process up to the prototype board was, so make it niche, do something that others haven't done before. Of course, you build on the work of others. You stand on top of the other's work. But find a niche. And also it helps to document. So Marcel started documenting this from the second board that you saw in the previous slide on hackaday.io. And that is really helpful. Because other people that might also be interested will spark more motivation in you to go further. OK, so that's about the prototyping stage. So that took quite a while. We had a prototype. And then some people said, well, that's interesting. Why don't you make this into a kit? Maybe other people are interested in this as well. And so that's what the main topic is of this. And it seemed to be a nice little bit of follow-up work to get this as a kit. And we spoke to a few kit builders that we knew about what to do, how to approach this. And they were quite helpful to us. And that's also the reason why I'm now telling you my lessons learned in the hope that you will benefit from what we have learned. Yeah, so this is the working prototype. I think the work from the first buying the TTL chips to this stage was about 30% of the work. And from this stage to actually having kits that we could sell was about 70% of the work. So that's something that some people do overlook. They think, ah, this is all nice. We have a prototype. We're ready. No, we are not even halfway. That doesn't mean that the rest of the process isn't fun, because it was quite interesting. But it's not done. It's quite a bit of work. So this was June 2017 prototype. And in February 2018, we had the complete kit that we could sell. So a PCB in a nice little box, all the components, a manual controller, et cetera. So this process of going from this to this was about, well, Marcel and I, two people. We spent at least two days a week on this for that amount of time. So you can imagine that's quite a bit of work. We settled on a way of working where we were being pragmatic. Let's not overthink it, but let's think about every step of the way to find quickly the best possible solution. For the kit, we wanted to be something that could still be a bit useful in today's world. But we tried to make it really retro in the 70s. But we were pragmatic. So for instance, it has VGA output, which was not really an option in the 1970s. And for me, VGA is quite normal, because I'm a bit of an older grumpy old hacker. But for people that live now, VGA is already old school. So that was nice in the middle. And we also were pragmatic, for instance, in the usage of the chips that we used. So we wanted it to be 70s, where we would be using very high power consuming chips that are also no longer to be found easily. And we settled for newer versions that are plug-in replacements, but are using a lot less power. And another example is the power itself. The thing needs to be powered. In the 70s and 80s, you would have a big hunking transformer on the unit. But we thought, well, this all runs on five volts. So why don't you use B power? We also thought about a mains adapter, but then you would have the problem of having different adapters for different parts of the world. And with the USB, it's much easier. But then there's still USB-C, USB-A, micro USB, mini USB. And here we thought, well, the micro USB can be obtained in a through-hole soldering variant. So that's this one. And the other ones that you would normally find on an end system are SMD soldering. SMD soldering is not that difficult, but I do remember for myself that the first time I did SMD soldering was a bit apprehensive. This could be hard. It's not that hard, but this could be an issue for potential buyers. And we settled for the mini USB. Another example, the prototype board has two chips, which are PROMs. So they contain, well, the gigatron works with 8-bit opcodes. And each one has one 8-bit data, 8 bits of data. And that's why there are two ROMs in the original gigatron. And we combine them into one. And you can buy those 16-bit PROMs. But I grew up, years ago, when I opened up my computers, they would all have these nice EPROMs with a little glass inside where you could erase them with UV light. And I really like that idea, so I wanted to get an EPROM in there. Of course, if people buy the gigatron and want to use an EEPROM that you can electronically erase, it's a plug-in replacement. But we settled for this. And also, the clock I found in the prototyping board, there is a clock made out of it. And I have to look, because I always mix this up. This is a crystal oscillator. And this is actually a little unit that contains several components. And I know from the old computers, they always have a crystal resonator. And this makes it a bit more complicated. You need a few more extra components. But I think it really adds to the retro feel of the system. So that was changed, as well, in the final design. One of the very important lessons in creating a kit is to make things as simple as possible. This was also a goal in the whole gigatron design process. But also for the kit, it's very important. Because less stuff in the kit means that it is less work to do for your packaging, less chances of errors in packaging and missing items. It's also less work for the people building the kit. And you will have a lower cost, because you have less components. And we tried to do as much as we could in software instead of hardware minimizing on the complexity. And that worked quite well. It did mean we had to do some concessions. So at first, Marcel was really keen on having really a super standout form factor for this box, for this gigatron. So on the left, you see a really nice, interesting shape of interconnected PCBs. And Marcel was thinking about a setup like this. So it's something that really catches your eye. But of course, this is incredibly complex to make, to design, and also error prone for people trying to build this. And we spent a lot of time looking for the right enclosure. And we talked to some people, also people from my hometown of Eindhoven, of the Crypto Museum. And they had a tip to go to people who make cigar boxes. In the Eindhoven region, there used to be a lot of cigar factories. There are still a few. And this comes from a cigar box factory. And yeah, it is a bit more dull. It's a rectangular shape, but it works well. And it still, I think, looks pretty classy. And the supplier was really helpful. He said, because we thought we need to have screws and hinges and stuff. And this guy was really super. He said, no, we can do this like this. So we can put a bit of plexiglass in there. And then it fits in there nicely. And you can just close the lid. And we'll put some wood in, and it will fit without using any screws at all. We just need to drill some holes in the back for the connectors. We also have some LEDs on top, but they do not need any holes because we have plexiglass on top. And it did mean that we decided not to have any buttons at all on the device because that would require additional drilling or anything else to make that work. So we changed the electronics. As you see on the top right, we added some electronics to make sure that the system will boot up nicely without having a reset button. And the reset can be done in software. And then we needed a controller because that was not included in the kit. And we had no idea what to use. I found some controllers on the internet in China somewhere. And this is a Famicom controller, also known as a NES controller. It's not completely the same. They come normally with a NES plug that you see on the left. But this plug, well, it's really hard to get the connectors for them. And they are hard to find, they're expensive. And, but I did find a supplier who had those with a standard DB9 connector, which is very common and very easy to get. And so we decided to own that one. So we bought it in China, just bought a few, and we started experimenting how this thing worked. So we put it on the oscilloscope and we found out that it's really a very simple shift register. So you give it a pulse and then you, to wake it up, then you give it a second pulse and each pulse you get one bit of button data out of it. And that was really great because it allowed us to hook it up to the Gigatron actually without any additional hardware. Why? Well, now I go into a little bit of technical detail. The Gigatron is outputting a bit-binging VGA signal and the VGA signal has a horizontal sync pulse at the end of every scan line. And then at the last line of the screen, it has a vertical sync pulse. And we use these to trigger the controller. So every vertical sync pulse, the controller is being read and then every horizontal line, it reads one bit out of the button state of the controller. And for that, we just needed to add some extra wires and fix the rest of the software, which was great. And also since we were working on this, we could also add LEDs with just one component, extra plus the LEDs and audio out with four extra components. The next big thing was to get this onto a PCB because we had the prototyping board, but of course we wanted to have a PCB. And this required learning key cat and getting all of this on there. And this is all, those traces are handmade. I mean, you can do some auto tracing but this is all made by hand to make sure that it looks nice but also that it is logical. So for instance, over here you see the algorithmic logic unit and you see there's four chips here on the outside and there as well. And those are 48 bits that it uses in the calculations for instance. So this is logically arranged but that took a lot of work. It took 10 weeks of work to get this to the stage of this board. And in this process, we also had to know which actual components we would be using in the kit because you need to have the holes in the right place, obviously. And that's where we also really tried to make sure that everything that you can put in the board can all be accidentally switched. So on the right you see two types of capacitors and they have different values so they should not be intermixed. So we selected from the vendor different capacitors that have the pins in a different spacing. So it becomes harder to accidentally swap them out. So these things were taken into account for the PCB. And we made a few and there were two mistakes still in there which we caught before we hooked it up. We fixed those and we turned it on and it actually worked and here you see it showing a picture of itself as a prototype board. I have to say we were a bit lucky because this computer runs at a very low frequency at 6.25 megahertz and at that clock speed, it's not really that important how the traces are laid out. If you work on high frequency systems that becomes really important. But for this system, it's really tolerant to making all kinds of rookie mistakes in the RF field. So that was good. And lessons we had from the people that we talked to from the other kid vendors, from the people that made PDP and from the crypto museum, the Enigma, was that we should minimize support effort because that takes up a lot of time. If you supply a kid and it's not really easy to build and people start asking questions, which of course they are entitled to do, but that takes up a lot of time. And it's better to spend the time upfront to make sure that everything is, everybody will be able to build it in one go without any issues. So that means making everything unambiguous. So that's why we use a different component so they cannot be intermixed. Anticipate everything that somebody could do wrong. And again, simplicity. So that was a lot of work. In the end, this prototype board, if you have really good eyes, you can see this version, already version 1738. And a few versions later, we had the final board. And here we did our best to make sure that everything is labeled as good as we can to avoid mistakes of people putting chips in the other way around, et cetera. Then I remember we also, Marcin and I had a discussion about sockets and that was really a controversial discussion. Should we supply chip sockets with the kid? So you solder in the socket and then you put in the IC into the socket. And well, you could say the pro is that it looks really neat. It looks professional, like a high quality product. And of course, if you ruin a chip, you can just easily replace it. So those are the pros. But it is more costly because you have to add them. You also need several types. So you have the mixing up problem again. And there's a high chance that people who are not accustomed to using those to insert the ICs in the wrong way, bending pins, and sometimes they bend to the inside and you put in the IC and you don't see that the pin is bent and it gives rise to all kinds of problems. So we were fearing all the support calls for those. And also the chip is there with a physical connection or a soldered connection. And that could also lead to issues, although maybe not very often. In the end, we settled that we needed a socket for the EEPROM because you want to take it out and erase it and reprogram it. Now you have ZIF sockets, zero insertion force sockets that have a little lever to easily extract them, but they're really big and bulky, or they are not so bulky, but really expensive. And we, in the end, we decided to use these cheap dual leaf contact sockets. They are a bit lower quality, but they have a bit more tolerance for not putting in the chip at the exact right spot. It will still fit. It makes it easier to insert the chip. For the RAM socket, well, we also put it in a socket to allow for further RAM upgrades. And here we use a more quality turn socket, but it's quite hard to put a chip into a 28 pin socket if you've never done that before. So you need to make sure that the pins are straight. So I decided to bend them and insert them in the socket for every kit. So the kit comes with a pre-inserted RAM into the socket. The problem is well that if you now solder any of the other ICs into the board and you do it the wrong way around, yeah, then you are really out of luck. You can desolder a complete chip, but if you do not have the experience in doing so, you probably will ruin the complete PCB. And we've had cases of people doing so, and we advised everybody who's soldered them in the wrong way to just cut all the pins and then solder out the pins one by one, which is quite easy, and then ask us for a new chip and put that one in. So that was the components. And then we had to have a manual. So I spent about a man month on creating this manual. That was a lot of work, but it includes a little bit of information about electronics, how all the components work. It includes information about soldering. How do you solder? And here you can see a picture of how you bend the ICs to make them fit into the board or in sockets. So the soldering instructions. And we also have it in a ring, so you can lay it flat, which makes it easier to have it open next to your bends while you are soldering. So we thought about all these things. And actually this is after the wooden case that we supply with the kit, but this was the most expensive part to make as well. I don't know, is there anybody here who has experience with a heat kit? Do it yourself, kids, by any chance. Anybody heard of those heat kits? No, those were a thing of the past. I was one person who knows them. A thing of the past, but they had really awesome manuals. As you can see here, I looked up this image after the fact, after I had written the manual, but you see this one also has soldering tips. So how do you solder? How do you recognize bad soldering? And that's really helpful because it saves a lot of work in fixing problems later if people make mistakes. So that led to this manual. And what also was really helpful is that we were doing this as a team. And for instance, Marcel has a lot of experience in software. I have less experience in software, but I have more experience in soldering, whereas Marcel didn't. And that really helps to get out of viewpoints. And for the soldering, for instance, Marcel told me that the first time he was soldering, he put in the IC on top of the PCB and also soldered on the top side. And I had not thought about people doing this because I had been soldering all my life and I know you soldered on the bottom side. But that's really useful information because now the manual can clearly warn people for such mistakes that people will probably make. And actually, quite a lot of kids were sold to people who did not have any prior soldering skills and they were still able to create this kit. So that was really good. Oh yeah, and also I'm used to, I've been soldering and doing electronics for a long time. So I know the resistor color coding table on the top of my head, but nowadays people use SMD resistors and they just have the number on there. So they are not used to using the colors. So we need to add the color codes, but I also made sure that in the steps where it describes how to actually put the components onto the board, that it does not only say you need to insert a 10K resistor, but also has the brown, black, et cetera, coding on there to make sure that people use the right components. Because not everybody likes to read, but people nowadays normally will go to YouTube to find information on how to do stuff, myself included, we thought, let's also make a series of videos. So the kind people of, again, the Crypto Museum in Eindhoven helped me out and provided me the space to do some soldering, and I made 13 videos of the whole process going through the whole Gigatron build. I checked yesterday and we're now at about 500 views. I'm not sure if this is people who are just like somebody to solder stuff and tell people about it, or if they are actually building a kit, but I think it was a nice addition. Then, source and parts. If you build a kit, you need to have the parts, and what's important for your prototype, you just buy the parts and you're done, but if you want to make a kit and sell it for a extended period of time, you need to make sure that you are able to get those components for all of that time. So we looked around to see what kind of people had these components and what their stock was, and we found out, well, the game controller, we had no idea how many there were in stock, so we took a bit of a gamble. The crystal was a bit of an issue and there is one resistant network that might be an issue, but that looked to be okay for stock, and then also we need the parts to be reliable. So we heard some really horror stories about cheap components being bought, even chips that were bought that turned out to be empty shells with pins. That does happen sometimes. So we went for sources like Mauser, Digi-Key, Farnel to be sure that we had the correct parts, except for three components. The eProm is no longer really in production, I think, I assume, but we found a source that had a lot of them and they looked to be new. I also did the swap test to see if the label was not printed on later, but it looked to be a genuine part, and for this component, it's not that critical because I need to program it anyway before it's being put in the kit, so we know that it is working. Now, the controller, that was a bit of an issue because we ordered a few of them, we sold some kits, it worked fine, and then we ordered our second batch, and then the controller was different. So there was different electronics inside, and to be technical, one was activated by an up flank and the other with a downward flank. So we had to rewrite the software, but we weren't able to make it work as smooth as the first controller. So all the people here and on the stream that have bought from that batch, my apologies for the controller not being as smooth as it could have been. And we had some problems with USB cables that were substandard, so they had a huge voltage drop over them, and if the voltage drops, it gets one no longer works reliably. And for the next batch, we switched to AWG-28, which is the measure of thickness of the copper inside of the cable, so thicker cables, and then all these problems were gone luckily. Now the first batch was 10 systems that we gave to beta testers to get valuable feedback about things that we could still improve, and then we had a 100, a batch of 100. And of course, the advantage of buying a bulk is that you get bulk discounts, so they start at 25 or 50 or 100 or even more. So we thought, well, let's hope we can sell 100. Actually, our goal was to outsell the Apple One, which sold 200 back in the day, so we hope you were going to make that. We had no idea if you were ever going to get there, but we thought we could take the risk of buying for 100 to drop the cost. So let's talk about the cost. If you build a kit and you are going to sell it, you need to determine what the end user price will be. How do you do that? Is it the cost of the components? No, it's more. And this is something that I didn't know at the time, but I took it from the URL here at the bottom and it says you take the cost of the stuff you bought for the kit, you do it times 1.66 to get your wholesale price, and then you do that again to get your retail price. And that might seem like a lot. Actually, we were below that figure, but we upped it substantially, but we need to pay taxes, we need to pay income tax on what we make. And what people sometimes forget is the amount of work that has gone into building the kit. So we did get a few people saying, well, I can buy these components for 80 euros. Why should I buy a double that amount? Well, that's because there's also a few many years or more than a many year in a few, maybe two, into design and then going from the design to the actual kit. And the process, when we got the kit, it still cost time to package all the kits and ship them out. So that was about 0.2 FTE continuously for the time we sold those kits for two years. So that is also to be taken into account and we need to reserve money for kids that are being returned or get broken or lost or whatever. So that figure is not that strange. If you're going to make a kit, you will also need to realize that you need to do an upfront investment because you need to buy your stuff before you get your income from selling them. And it is nice to get a legal entity to make sure that everything is in order with your local laws. I would also advise to discuss the financial with the partners you're working with on the kit. First of all, I would advise to work with a partner because it's more fun to do things in a group, but also because you can learn from each other that really worked well with Marcel and me, but also discuss the financials. What if somebody accidentally needs to stop with the project, which unfortunately Marcel had to. So it's good to make agreements upfront on what happens next and also talk about the amount of work that each of the team puts into the kit and how you then share the earnings in a fair way. I'd also like to talk a bit about kit packaging because that's quite a lot of work as well. So, as I said, we made sure that all the components were different and not so easy to mix up. So we could put all these standard components into a little anti-static bag, and that's that. And I did that using the rule, check everything twice. So I would make, for each of the components that need to go into the bag, I made little piles of maybe 10 or 20 of these components. So I count them, then I take the bag and put in all the components that need to go in there. And once all the bags are done, then there should be no components left. And if there's a resistor left, then I know that one of the bags will be missing a resistor and I have a lot of work in finding out which one. And then I also needed to count all the ICs, the chips that go in there. So I used a piece of anti-static foam on which to put all the chips. So I would open up a tube with the chip type A, put in the chips, then next one, next one. And this gives a really easy visual clue of whether the kit is complete or not. And again, check them twice. Then the bag, the ICs, the PCB, all other stuff went into the little box. And again, I would make little piles, collect them and check them, check if all the piles were empty at the end. Yeah, this is a kit that came back after a round trip over the world. So some kits, I bought a kit last a few months ago which was really nicely packaged. So it was specially purpose made foam with the indents where all the components went into and that was really nice because if you shake it well, it still stays in place. This foam does hold the chips quite well and in most cases, it will work quite well but if it is really handled in a bad way by the local postal office, this is what you get. But this was really incidental. So it was for us, it was good enough. And then in the end, the box with the components, the manual and the controller were put in a envelope. And again, we would do that twice. And if I were done, then forget to put a manual in one of the boxes and have one left at the end. Yeah, it would be a bit harder because I would need to open all of the envelopes to see which one was missing. But of course, you can then weigh all your envelopes to see which one is lighter so you know where you forgot the manual. Now, we sold the kit for two years and we met our goal of selling more than 200. We actually sold over a thousand. And as I said, it cost about 0.2 FTE during that time to do all the packaging of all the kits. And people have asked, couldn't you find some cheap labor to do that for you? Well, we didn't actually trust others with that. I'm not sure if that is the right decision. Well, actually, I think my opinion is it was the right decision because if there's a mistake, the amount of work that it takes to solve that problem, yeah, that's quite an investment in time. We've also got questions about pre-built kits. Can we provide them? And the short answer is no. If you decide to build a kit, it best is to sell them in kit form because then a lot of rules and regulations do not apply. If you build a ready-built electronics device, you will need to have CE and FCC markings on there and that's a really costly, long, elaborate process. In the middle, you see a test I did with Mark Siemens, who had a nice device to check the RF emanations from the device and it's horrible. So this will never pass to any FCC test. And yeah, but selling as a kit, no problem with that at all. For shipping, you can shop around for a good bulk shipping person, but there is no one-size-fits-all. You will, in the end, have to deal with different shipping options. If people live in Manila, they cannot rely on the local post office. You will need to use DHL or FedEx or something else to be able to send over there. And insurance, it would be nice, but we decided not to have insurance. In the beginning, we did have insurance and then if your package gets lost, you can apply for the insurance, but then they will say, well, only if we reimburse you the cost of the components in the kit, not the amount of money you sold it for. And please show me all the paperwork that explains how much the components were worth. So it's a lot of hassle and it's actually cheaper to just have these few parcels be lost and not pay for all this insurance. And yeah, in the end, maybe a portion of a percent had an issue. And some places are really hard to send packages to, so there was one sent to Brian in the Falklands and that one did never arrive, unfortunately. We also try to exceed customer expectations. So make something that people are really happy with and they are happy from getting something that is better than they expected to be getting. So that's why we spend so much time on the manual, make it super, make the support better than expected. So they're happy with the whole experience. And support we did via email, but we also tried to do a lot for people to do self help. So we had websites on the internet with all kinds of guides, how to measure voltages to pinpoint a problem. And quite early on, we made a forum. And we had very limited issues with the parts that we got in, only with the controller problem I told you about. And once, believe it or not, we had a customer saying, there were the wrong resistors in my package. Well, no worry because I'm an electronics geek so I have enough spare ones. And then somebody else said, I'm also missing resistors and it turned out I had swapped them in the two different bags. And of course, my little piles were all down to zero. So I didn't notice that I had swapped resistors in two packages, but that was the only mistake I made in two years, so that's fine. Marketing, we did a little bit. First of all, we used a really nice white box to send them in to get that first experience of, ah, this is a nice kit. But of course, we also need to do some marketing via vloggers and we decided to send a kit to the 8-bit guy who has a blog about 8-bit computing and we already knew in advance that he would not like the fact that it would not run basic. Actually, the current version does run basic, but the first version didn't and he also commented about it, but he did a really nice review and that video now has almost two and a half million hits. So that was really great. And we also sent one to the EEV blog, which is an electronics guy, and this was cool because he got the kit and he opened it up on this channel and he saw the manual, said that's nice, do it away, and then he just started, he saw on the PCB what should be on there and he just started soldering all the stuff on there without any manual at all and he got it working from the first try, so that was really great. And he did a live soldering session video and he also did some follow-up videos about, ah, this is a nice system to show you the difference between two-layer and four-layer PCBs. So together with us, he made a four-layer PCB and he did some measurements to see what the differences are and he also did a video about the caps that are there, well, the bypass capacitors. So that was cool. And we sold quite a few. In the first month, we sold to 25 countries. We sold to companies like Microsoft, Apple, Google, Facebook, IBM. Well, we sold to Apple, but I don't think we sold to Steve Wozniak, which was our super, super goal. We also created this community on form.gigatron.io and then it took off a little bit and that was really great to see if you have your brainchild come to life with people starting to use it and experiment themselves. And after one month, we had some additional software. We had Sprites. We had people doing hardware like RAM extensions. We had somebody who built a really fast version using special fast TTL chips that would run at quite a large speed. But of course, then the screen is reduced. We had things like this, like a special breadboard, sorry, breadboard PCB with everything in a very small form factor with more ROM and four-layer PCB. We had software like a C compiler. People are working on fourth for the Gigatron. We have video repeaters. We have all kinds of other stuff. And in the end, it's crazy what people have written. So on the right, yeah, that's actually coming from a system without a microprocessor with just under 1,000 logic gates doing the VGA as well. So that was cool. Now, issues, there weren't many. One issue we were worried about in the beginning was what if somebody takes our RDR and just clones the PCB? Copycats. And it was only after one and a half years that we found on eBay this Gigatron from Russia. And I was curious, most of us furious. I was curious, I ordered one, I had a look, and this thing does not work at all. It is really not a working copy, but it was interesting to see and there were only maybe 10 sold and that was it. So we didn't have any problems with that. We did expect that it would be a nice kit for educational institutes, but in fact, yeah, we didn't sell, well, we sold some to universities and such, but not that many as we would have expected. Another issue that we had, there was one guy who in the beginning asked for schematics. He was also working on a CPU for himself, which we explained were in the manual and then suddenly he became bad-mouthing us on all kinds of occasions. We just ignored him and so did the other people on the internet. So that was also controllable. Now, after two years, the project ended. At that time, Marcel was terminally ill. We spoke about this project. He had actually plans to continue and maybe port CPM to it and do some other stuff. And I said, well, actually it was really fun to create, to make the design and also make the kit, but putting resistors in little bags for one day a week, yeah, that's really work. That's not hacking. So for me, there was not much fun in there. Well, there's still fun in following the community and what people are building, but not so much in creating actual kits. So everything was then open sourced. You can find everything online and a few companies have started to sell kits as well. They don't have the manual, which you can download from the internet and they don't have the nice little wooden box and some also do not include the controller, but it is enough to get a working gigatron. So important lessons that we've learned is it really pays off to have a minimalistic design. That makes it easier to create the kit. It makes it cheaper. It makes it less error prone. The problematic approach, we tried to have a set of rules to stick by, like making it a 70s, 80s error machine, but where needed, we would deviate a little bit where we felt it was still right and tried to put a lot of effort, not in support after you've sold the kits, but get that time spent in advance to make sure that people do not need the support. Make sure that you have a team because two, no more than one, so that really helps in making the product better. And we were lucky that we in the beginning made sure that we were able to update the software later and add functionality to it. And of course, most important is to just have fun doing this. It's a really fun project. You learn a lot and stop when it becomes work. And the greatest achievement I think we've got is that Indiana One was sent to Steve Wozniak. So I know he has one. He was impressed by the chip account on the system and he actually did sign a gigatron that I have here. So to Infinity and Beyond, thank you very much for your attention. And maybe I have 30 seconds for questions or... Yeah, go for it. You've got two minutes. So yeah, if there's any questions for GigaWaltz, please quickly head up to the microphones in the center and be brief. Yes, on the front microphone. Was... A little closer, please. Was sending kits to these guys on YouTube really all the marketing you did? Like all the public relations, whatever you would call it? What we did was... Well, most of the marketing was on hackaday.io. So that created already a bit of a buzz. And then the online marketing we did was only create a forum that we didn't advertise any further and the two YouTube videos. All right, thank you. You're welcome. Yes. Hi. On the question of how to source things as two source parts, you mentioned to our resistor network. Yeah. If that was such a problem, wouldn't it have been easier to just use discrete resistors there? Yes, so that was our backup plan. So we thought if the resistor network is no longer available, we'll just get some resistors. And actually, they have now become really hard to source. So one of the suppliers that is now selling GigaTron kits is using that approach. Okay, thank you. So it was a good question. Thank you. Can you tell something about how people program the GigaTron? Yeah, there are several ways to program it. We have a, well, you can program it in machine language, but it is really hard because you have to keep track of all the timing to make sure that the VGA signal is still correct. So nobody does that. Then we have something called GigaTron command language, GCL, that's rather low level still, but good enough to be used to create some games or other things. And now, of course, we also have the little C compiler. We have basic. So there are a few ways to program it. But do you need to refresh the EEPROM? Yeah, so basically you need to refresh the EEPROM, but there are now hardware extensions. So this is a hardware extension that also includes an SD card where you can just put your files on the SD card and run them from there. So that's not part of the original design. That's something that came out of our community. Nice. Thank you so much for your questions. I think all the remains is to join me in thanking GigaWalt for a fantastic talk. Thank you.