 Welcome to Hackware for September 2022, the first in-person Hackware after God was born. Thanks, everybody, for coming down. And welcome to the new Hackerspace. It's my first time here as well. So if you want to know where the toilet is, I don't know. Not someone else. So in case anybody doesn't know what the square is, it was originally formed on Facebook. It's just a group of people who are interested in electronics, hardware, IoT, and general kind of hardware-y things fell out of a bunch of interest groups back in the day with around Arduino and Raspberry Pi. So we are mostly chatting on Facebook, although that's kind of gotten down a little bit. There's an official media.com thing there where we actually organize the physical media. This is a team who is organizing it. It's kind of a very loosely organized team. So you'll see different faces every time potentially. And if you're interested in helping out to organize this, talk to us. We'll be more than happy to have more help. So now we'll quickly talk about a few upcoming things that are happening in the community. The first thing I'm helping out is a Singapore AAB challenge, which is a student competition for underwater robots. I can quickly show you. It's a competition for students who build these kind of underwater robots. The actual competition is open to public, which means you can watch. The competition is only open for students, and all the registration is done. But if you want to watch, you can watch. It's over three days, so it's quite long. Usually the Sunday of the competition is where most of the action is. But it's still fun to come and geek out about robots and fun, nerdy, hackery things. Another thing we've already talked about, there's Geekcamp Singapore that's happening. And if you're interested in giving a talk, you can go to Geekcamp. The call for papers is, if there are any other related user groups that are in their meetups and want to talk about it right now, raise your hand. You can mention it, and then hopefully we can have a bigger community of hardware communities in Singapore. Anyone? Want to start a new hardware IoT electronic group or something like that on those lines? We used to have lots. We used to have a robot operating system specific group. I think there's still some people doing that. There was an FPGA group, but pandemic things just disappeared. Hackware has a telegram. If you're interested in talking like the Facebook, there's a telegram chat, which is a little bit more active, I think, than the Facebook. A different group of people on Facebook and telegram. But if you're interested in talking more about hardware things, you can join the Hackware telegram chat. Thanks to Hackerspace for the venue. Quick show of hands. Who is the first time at Hackerspace? Or any Hackerspace in Singapore? First time? All right. I'm here at this iteration of Hackerspace. I was there the previous one. This is my first time here. In case you don't know what Hackerspace is, it's a community space. The idea is to come in and go out and use it for doing community events, putting on projects. You can rent this back in the office area. There's a hardware room over here. Unfortunately, I have no idea what's inside. It might kill us. Oh, OK. Give us a tour of the Hackerspace data. It might be a movie trap for you. Thanks to engineers at SG for live streaming the whole event, so the events also live stream and capture. If you guys don't know engineers at SG, you have to go to the website. It is amazing. Michael and his group have been capturing the videos of live events when it used to happen back in the day before the pandemic for, I don't know, 10 years. And it's an amazing archive of all awesome talks that have ever happened. And many, many other meetups in Singapore. And the YouTube channel is awesome. So I'm going to check it out. And a big thanks to them for covering this and this. Amazingly hard work for so many years. So thanks, Michael. We have three talks. The first one is a joke talk by me. I put it in because nobody wanted to sign up. And I decided I would just talk about something random. So don't expect much on that one. Talk is by Michael, not that Michael, this Michael, about a robot. Oh, hello. That's over here, I think. And then the third talks by Claudio. That's about an accidental projector. So I will switch over to my presentation and I'll start the talk. Let's see. All right. No worries. All good? By PCB tricks, I learned from staring at people's design. The pandemic was long and I was bored sometimes. So I was staring at people's PCBs. So I learned a few interesting things and I thought it would be fun to share these. Before we start, who am I? I am not a hardware designer. I am a software guy. I have almost no PCB design experience. I've built two PCBs in my life. And they were both tutorials. So I have really no experience. I love to let out magic smoke from PCBs and electronics. Usually, that's something that happens very commonly. Do you know what is bodge life? Bodge life is, again, a joke, a hashtag about how all of your PCBs have bodges. And every single PCB that I ever touch ends up having a bodge. So I'm really not at all a person who should be giving a talk about this. But since there was nobody talking at the meetup, I decided I'll try it out. So I'm going to talk about something fun. But it is actually really fun what you can see on Twitter. And it is really exciting. And I think that is what really got me really into electronics. Recently, it has been some really fun things that you can see on Twitter. I mean, look at this PCB. Doesn't it look really amazing and beautiful? And I think electronic Twitter has been amazing. For whatever reasons, there's a lot of really cool people who do electronics on Twitter. And they do really fun stuff. And they do, they share their designs. And it's really good to learn. And as well as to just all go at really pretty PCBs, like this. But if you're not following, if you're not on Twitter, if you're not, if you are on Twitter, if you want to follow electronics folks on Twitter, there's a bunch, and they are really fun. They do a lot of really interesting discussions. They rant about part shortages and the usual stuff. And it's awesome. Open Source hardware has been a really great movement for this as well. People have been sharing designs a lot more openly. And I think that has really gotten people like me who are at the fringe, being able to experiment and play around with hardware and really understand how things work. KiCad, if you don't know, is an open source EDA tool. This is for designing schematics and making PCBs. And it's awesome. And I love trying to make stuff. I try to make stuff. It doesn't work. I give up half the time. But it's been fun. And I think that's where most of my lessons come from is some of these things. So warning, if you're a serious electronics person, the tips I'm going to give you are extremely basic. All of them have trade-offs as with anything with engineering. And it's all rabbit hole material. I'm sure with people who are experts, we can talk about any of them. One of the tips for like two hours, four hours, 10 hours, because there's always nuances and small little details for everything that we do. So let's go. This one was interesting. When I first heard about this, I was like, what are you doing? There's this concept of a zero-ohm resistor that people input on PCBs. And I'm like, why do you even want a zero-ohm resistor? Isn't that just a piece of wire? Like, why do you need to put a resistor? Why do you need pads on this? But it's actually pretty useful. So standard electronic circuit is a regulator that takes in some input power and generates 3.3 volts. And if you put a zero-ohm resistor here, you can isolate the rest of your circuit from your power source. So you could remove the zero-ohm resistor and then sort of test your regulator in isolation. And then if it works, if it doesn't work, fix it. If you need bodges, I always do. Bodge it and fix whatever things you have broken, and then put back the resistor. So you know that the rest of your circuit isn't affecting the regulator. So it's a very good way to isolate two parts of your circuit, not only from the power side. The power is something that you do very commonly. Also common to use this too. So if you see this JP5 and JP4, so these are same things. They're actually shown as jumpers, but they're actually just zero-ohm resistors. So you could isolate subsection. So let's say I don't want to test the e-ing side of things first. I don't want to test the Laura side of things first. I want to test the rest of my circuit. I want to make sure everything else is stable before I bring these up. Maybe they're causing problems. So again, zero-ohm resistors, put them on your board, remove them, these hold them, test your circuit out, and when you're done, you can put them back in. So a very cool technique to isolate parts of your circuit from the rest. Another nice thing that I've learned, so a resistor footprint on a layout would look like this. I'm going to have to walk over for this one. You can actually take, so if you pop the resistor out, if you don't solder a resistor, you can actually solder a wire here and a wire here and put it across an emitter or a multimeter in the current board. And you can measure how much current is going through this. So especially if this is a power supply and you want to measure how much your board is drawing or certain part of the board is drawing, it's a very easy trick to put a multimeter or an emitter in the middle, which is something normally very hard to do with the cut traces and sort of solder on one side and on the other side, it gets very awkward. But if you have a zero-ohm resistor there, just pop it and then put across the pair. So another very nifty way to go for it. You can see the difference. You can, you can, you can. But if you're assembling large boards and you have lots of parts, you can actually get zero-ohm resistors so you can get them picked and placed or whatever thing. So it's, you can, you can put a wire. You can put what you need, actually. It is. Well, you can use it for debugging. I find it much more useful for debugging because of all the, these things. But yeah, you could, you could put a wire. I have even seen people with very fancy layout pads. I'll show you later. Where you don't even need to put a resistor, you can just put like a, a big block of solder. And that works too. So, but there's this super interesting thing. Next one, I2C address pull-up and pull-down. This one's something that bites me all the time. Lots of times I2C peripherals tend to have this stupid thing about being able to have different addresses and then you have generally two pins that you have to either pull up or pull down or not or refloating. And then depending on that, they will have a different address that means to address them. This is usually done if you want to have two of the same chips on the same I2C bus. I2C is a bus protocol. So you can address them differently so you can actually talk to both on the same bus. The trick I have that I see a lot of people using is to have footprints for both pull-ups and pull-downs. So during many PCB design time, you can have all the four resistors in the pads. And then when you actually populate your board, you can design which one you want to populate. Very useful because you always make mistakes. I always make mistakes with this. And I'm like, oh, I'm going to design for blah, blah, blah. And then when you actually start writing something, realize it's a crap. I designed for the wrong one. So having the footprints there, it's great. You don't have to populate all of the four 1K resistors. Just put the ones that you need. So if you just want one pull-up, one pull-down, just put the two and then need the rest unpopulated. So it's a very easy way to sort of have all your eggs in the basket at PCB manufacturing time and then you can figure out what you want to do at actual building time. Breaking our MCU pins, this one has bitten me the most, like so many times. Microcontrollers have like a zillion pins and usually you're like, oh, I'm only going to need like these two I2C buses and I'm going to need this and that and then that's all. And then the rest of the pins, I'm just going to not do anything with them because I'm too lazy to lay out things. And then you realize that two days later that, oh crap. If you look at the pin-muxing diagram or the table, which is extremely confusing, and the first time I read it, I always get it wrong, that the exact pin I need, the exact peripheral I need, like I need an I2C, it can't be routed on the pins that I actually have out. Some of the modern microcontrollers, thankfully allow a lot better routing, but all the ones are crap at this. So I like what the folks at Teensies did. Look at the back of the thing. They have like all this tiny, tiny little pull out. So they pull out almost literally like every single pin there is on the microcontroller. So in case you mess up, which I do all the time, you will at least have the pin pulled out so you can solder a wire to it and then sort of, you know, do something about it. Maybe bought something else. But otherwise, especially if you're a BG or one of these really tight ICs, why is it a pin from that point forward? I mean, even if it's not a BGA, it's a very tiny thing, soldering to those pins is impossible. I mean, you potentially could someone like, someone who has much experience could, I can't. It's just beyond me. So I think it's just easier to pull them out, especially if you're doing a deaf board, and if you're not really restricted in space, put something really, really tiny. Even if it's a small little test point, you can potentially solder to that as compared to soldering to this tiny, tiny pins on the board that you forgot to pull out. So just break out every single MCU pin, especially if you're building a first board, just break out everything. Number four, reverse polarity protection. This one, something that I heard from experts and they were like, oh, you should always have reverse polarity protection for your circuit. And I was like, yeah, sure, whatever, right? And then of course I do this and you connect the red and the black invertly and then puff goes your circuits, like crap. In fact, happened last week at work. My colleague was installing something and then it didn't work and we're like, why is it working? And we look open like, oh crap, we inverted it. Thankfully, we had reverse polarity protection, which means doing this isn't blow up a circuit, it just doesn't make the circuit work. Fine, that's at least a safe default. There's many easy ways to do it. They usually require a couple of chips and resistors, but look up, I think, what's his name? Great Scott did like an entire video of like going through all the different ways of doing this. There are some benefits pros and cons. Now look this up and there's always something interesting. There's always, otherwise there's always a yellow way of doing it, which is just put a reverse diode across it and then your power supply can go burn if it needs to. Kidding, don't do this, just don't do this. I'm kidding. Watch the video and try to figure out which is a way that works for your circuit, but I think this is a very important thing to add to anything you design. Put it back once more together. So I think this type of connection that you take off, it won't close with you. No, no, no, no. So this one, if you need some kind of a circuit like this on your board to do it. This is just me doing stupid things, which I tend to do a lot. Just, yeah. So you need a circuit like this on your board, but they're quite easy to do. Quite, quite, in fact, you can even get modules that do this for you. And this video tells you about the different ways you can do them. Or you can do them this way, but not recommend it. It will blow up your power supply. Last one, LEDs everywhere, the sun's straight forward. Just put as many LEDs as you can. But I can, I think what I've shown here is a couple of places where I think LEDs are good. If you have an input power, I think an LED on the input power is very useful because you want to know whether whatever is inputting actually giving you power. Very useful to have this for quick debugging. If you have a power regulator, then on the output of the power regulator, having an LED that actually shows you whether your power regulator is working and whether your 3.3 volt or whatever rail is up and happy, also good to have. If something else shorts this rail, this LED will go away, so it's very easy to know what happened, or what went wrong. Or at least you know that, oh, my 3.3 volt wins it. And I always recommend at least one user controllable LED on the microcontroller. Because when nothing works in a microcontroller, the least you want to do is say, okay, sanity check. Can I toggle the LED from my software? Is it working? If it's working, then at least, you know, you're able to flash your software or whatever. So I think having at least one user controllable LED on the microcontroller is recommended. The other thing I learned, another quick trick is if you are, so one of these people who is here having too many LEDs is, oh, it's drawing too much power. It makes the circuit draw, and if it's battery operated, especially, it makes it draw too much power. One thing you can do is play with the resistor value. So maybe initially you have them low resistances, so the LEDs are nice and bright, and then if you wanna like, you know, deploy it at some point, you can just increase the resistances and make them very thin, or even better, just not populate these two and the LEDs go both. Yeah. That is a reasonable transition question. Yeah. How many parts are working? Exactly. Pull the LEDs up, pull the resistors or whatever. So easy things to do. So that's five little tricks that I learned just looking at people's designs. Thank you for listening to my talk. I have a quick announcement. Subnero, that's where I work, is hiring an embedded software engineer. So if you'd like to do embedded stuff, Linux ARM, microcontroller stuff, and if you're in Singapore, let me know, we can talk after the meetup as well. We do underwater comms, so we do underwater communications. So we make the underwater wifi. But happy to talk about it in detail as well. That's all for now for me. So thanks. Voschens, yeah? Sure, what? I would say the community has got to be with us. I, yes, I am into audio stuff. I kind of do, yeah. Are you making the hype up or? Not very high. Not high-five kind of audio, but just random audio stuff. Yes. I made the assumption that your company is yourself. No. Well, I'm not, but yes, we have people in the community. I don't think Leon's here today. We have a friend who does it. You do a lot of manufacturers online and you specify for that, so you can pull out, right? They, I imagine that you want to pull out because you don't want it to hang in and then it can be unwanted. Or you might want to use it tomorrow. Right. You might want to use it tomorrow. If you have a lot of manufacturers that can actually, so what happens is that they allow you to specify multiple soldering points. So one of them would be linked to a common ground that you want to do. So you don't really have to damage it. It might possibly be worse for damage. If you want to, you can just ask them to do that for you and then you just, I guess you could just solder it to a common ground or both at night. Yep. Yeah, and then you can solder it and then you do what? Yep. Thank you very much. If you're familiar, I've seen that you've got the two same as ever hatches, but actually as fair with the trace in place. Yeah, yeah, yeah. And you can cut it. The traces. The traces. Yeah. I know a lot of Adafruit modules tend to have that kind of a design. Yeah, that's true. Yeah. Just to give you a general, these are all essentially the sort of test point entries and configuration options and so on. Yeah, it is, it is. Any extra stuff to make a change in runtime that you might want to come up with? Exactly. So that's a test of what's going to be done. So that's what you need to do. Cool. I have a question for you. Julia, is that the software computing languages? This one, is it zero? Yeah. No, we do underwater communications. Okay, so for the next room, they are packed by... Oh, Julia! It is, yeah, it is a computing, numerical computing language. Oh, is it used? Am I wrong for data analytics? It can be used for data analytics. It can be used for any kind of numerical computing. Okay. We can talk about Julia later. I can talk to Kaus Kammelma about Julia. All right, yes. So how do you actually do this before you type in the... No, as in there are, but most of them wouldn't bother making PCBs for you unless you are ordering in large amounts locally or like super complex ones. So most of the time send to China. So things like PCBWay or JLCPCB... So no one's doing PCBs at all? There are, there are. Actually, the people... Leon's not here today, so we have one of our friends. He actually gave a bunch of talks at Hackware before. He actually does PCBs at home. He has tried a bunch of iterations. In fact, he gave a bunch of talks about his various iterations of making PCBs at home. People have tried that and... So I think Leon was using salt and hydrogen peroxide. But even that, the getting access to high concentration hydrogen peroxide is online in Singapore. So he gets information. So you can, but it's just like, it's industrial. So you got to go to like... And also, with hydrogen peroxide, it takes time. Compared to the product. So it's a bit slower. But also, you're not going to get anywhere the kind of precision that you get with factories. And it's so cheap to get them made in China these days. It's like $20 for small ones. It's more about the activation. It is, it is. I've seen people with CNC-based PCB-mining machines. So those I have, I know a few people have those. So that, that, that's, but that only, we can only do two-sided. So you kind of remit it on that. The accuracy of all things. Of course. Thanks. All right, next speaker, Michael. All right, let me just, oh, okay, let me be right up. So let me just, and how they're doing, you know. Okay, cool. Okay, let me test. All right, cool. All right, hi guys. My name is Michael. And just want to share about this project that a bunch of us have been working for for over half a year now. So basically what we're trying to do is we're trying to build a decentralized network of delivery robots. So these are sidewalk robots that are not meant for, to be on the road, but these are just on a sidewalk. And so maybe let me just show you some, like, live stream that I did. Let me see. Oh, wait, wait, is it a clip playing? Sorry. Okay. Yeah, so, so you can see my controller, so I'm actually controlling this robot. Actually, we brought the fiscal robot later on, I'll bring it up, but essentially this robot has a couple of things. It's got a front camera, rear camera, it's got GPS, it's got IMU. And effectively it's, I think this version is not fully WebRTC yet, but basically, obviously there's a way to control the robot. And as you can see, I'm just remotely controlling it. And this is all done from my browser. So, yeah, and I think this trip I happen to be going from, I think, issue into Woodlands or something like that, yeah. So, let me show you another one. So this is from my perspective. Maybe I'll show you another perspective, so that's kind of from, so we put in a 360 camera on top of the robot, so you can kind of see, we also bought the real thing here. Let me bring it up so you guys can see as well. Yeah. Feel free to come and take a closer look after this. Oh, sorry, I got, okay, okay. So this is not this exact robot, but this is, we have a couple of these now. I think this is still using, running on jesson nano. This is already Raspberry Pi, yeah. So maybe I should, okay, maybe I'll talk about this video first. So, here we're trying to cross the road. I guess, yeah, we have to speed up, because actually right now, this current version of the robot is pretty slow. The maximum speed is something like three kilometers per hour. So it's almost like a slow walk. I think the goal is to get something that's around five kilometers per hour. Yeah, so it's kind of like a fast walk. And this is quite interesting. So we actually drove the boat into a, I think kind of a hawker center and pick up some food, yeah. So I actually wanted to let you guys try driving it now, like in real time, but actually the other robot, like our other teammates driving right now, as of now at Marina Bay Sands, they're picking up some bubble tea there. So I, unfortunately, I can't let you guys try. But yeah, I mean, this is one that we recorded this, obviously. So I think in this case, what happened was we, we ordered on grab food, pick up, and there's a speaker and microphone, and the guy driving just, hey, you know, this is my order number, one, two, three, you know, like just, just put it in. Yeah, so, so, I mean, this is the basic concept, right? So maybe I step back a little bit and talk about like the genesis of this whole thing. So last year, I was actually still working for another startup, like at the time I was doing like an edge-to-tech startup. And me and my two younger brothers, so one of them, he's been, he's based in China. He's been based in China. He's basically a China man now, like more than 10 years. He's kind of architect by training, but he does a lot of like, let's say physical design, a furniture design, a little bit of product design. So he's based in China, but my youngest brother, he's actually, right now he actually works in Tesla in autopilot. So I will consider him a really well-class self-driving guy, yeah. So really the genius and the family. So, I mean, I'm in Singapore obviously. So three of us, we thought, hey, you know, three of us are all very geeky. We thought, okay, let's do a hobby project so that, you know, three brothers, you know, we can kind of keep in touch with each other and whatnot. And so, so it really just started as a weekend project. We start hacking something together. I think the first robot we got was something or more almost like ready-made on Taobao. We bought it for like, I think $1,000 plus US dollar. And then we put in just a nano and then we run Ross on it then. And I think the first time we tried out was, the robot was in Mountain View. So my brother in Mountain View has a robot. So I was driving it. So that was the first time I tried out. But as you can imagine, it's just full of, I mean, we weren't doing it seriously anyway. But then towards the end of last year, we started like thinking quite seriously about doing it properly as a like, you know, almost like a full-time thing. And one of the, there are two, I would say, main thing that happened that kind of was certainly me and my second brother. So actually me and my second brother or both of us quit our jobs to do this full-time. Our youngest brother is still in Tesla, good for him. So two development, like I said, so the first one I think it was, there was this other like sidewalk delivery robot startup called Coco, COCO. It's actually kind of a spin-up from, I think, UCLA. They were also doing something like this. And then when I try to read up, what happened was they spin out, they raised maybe a seat round, $5 million, you know, like U.S. seat round is $5 million, U.S. dollar. But then quite quickly after that, I think they raised like $35 million or something like that. And decent money for series A, but more importantly, it was actually backed by this guy, Sam Ordman, you know, who is the, now he's the CEO of OpenAI, used to be the president of Y Combinator. So I thought, hey, you know that there's something brewing here. And obviously they are not the only sidewalk delivery robot startup who has raised, there are other startups who have actually raised more. But that's the one that made me sit up and think, hey, maybe there's something here, like there's some serious backer, backing this kind of idea. So, and I really love doing robot. I mean, even though to be frank, I'm not a real engineer. So I'm the guy who basically tried to pull in a bunch of kicks to actually do it. I'm just a talker actually. But that was the one thing that made us think seriously about that. The other thing was, I started chatting with a friend and he was telling me about this project called Helium. Anyone of you know Helium? It's a crypto project. Anyone have heard of Helium? Oh, okay. Okay. So, yeah, maybe I can show you guys Helium. Helium is this, yeah, pretty mine. I mean, to me, I just, the first time I read about this, I was really quite mind blown to be frank. So it's a blockchain thing. It's a crypto thing. They run their own blockchain, but I think they are gonna switch to Solana soon. But effectively they are a, what they call a decentralized network of wireless network. And it's, initially they were on a Laura Wendt, yeah. So, and I think that they started as a startup, like years ago, I think 2013, but for the first four or five years, they didn't have any much traction because they have a classic, like chicken and egg issue, right? Where for this network to work, they, you kind of need to form a mesh network in order for you to provide like consistent network coverage. So you have a classic chicken and egg thing, like why would someone buy the router from them? If like, you know, the network doesn't exist already. And then I think almost like a Hail Mary, last move kind of thing, like in 2018 or 2019, they went and do this blockchain thing where the basic premise is you buy the router from them, you just switch it on and they have this thing called proof of coverage where if there's another hotspot nearby that can see you and you also see the other guy and you kind of verify each other, right? In a kind of trustless manner, then they will give you this token, right? They call it HNT. Now, we can all argue why that token has value, but somehow this token has value, okay? And so they kind of printed, you know, a kind of currency in, from thin air, but because of that, right? What is super amazing to me is like, you know, let's see, today they have, like from almost no one buying their hotspot within about three years, they have now closed a million hotspots around the world. And if I just, let's say, swimming to Singapore, I think Singapore is pretty well covered. Wow. Yeah, so you can see, let's say this, but I don't know, like where we are now, probably there's a, probably like four or five, like this hotspot is already on. One thing about this lower end is that usually that the coverage distance is quite wide. So you probably only need like four or five, it's decent coverage already. You don't really need more than four, like 10 is probably too much. Yeah. But again, what amazed me is that they, by introducing this so-called tokenomics, they managed to incentivize almost a million people around the world. And this hotspot sell for about, say, 4,500 US dollar. So effectively they, it's almost like running a Kickstarter program where they managed to raise like 2,300 million, right? Because almost a million people pay for 4,500. That's what I don't, I mean, I think at least a few hundred million, right? So they didn't have to raise their flows, 4,500 million to build the thing themselves. But more importantly, they managed to get people really around the world, like all the key gateway cities around the world. You will find pretty good coverage, like big cities like New York, London and whatnot. So I just thought that it was a very interesting, like to be frank, I'm not a crypto native at all. Like I haven't really bought Bitcoin still to do today, but I thought that this way, I mean, clearly they've shown that they have managed to create scale at a very, very fast pace by throwing this crypto thing into the mix. And it, but it's the whole thing is anchored by these IOT devices. So I thought, hey, you know that that could be something interesting here because like the way we were thinking about this delivery robot, maybe I should talk about like some of the so-called, I guess product philosophy about like what we're trying to do here. Now, there are a lot of this delivery robot startups around the world. I think Singapore couple of couple, you know, all the major countries will have a ton and some of them are very well-backed. What I realized is that most of them tend to focus very much on self-driving, right? Which is obviously my youngest brother is, to me, he's a world-class expert, right? So, and what ends up happening though, is if you try to do the self-driving, you know, deep learning thing, you end up having a pretty expensive unit because you need to load up with LiDAR, radar, you know, a bunch of sensors and then correspondingly, your computer is going to be very expensive, especially now, right? Justin Nano used to be, I don't know, the hundred bucks now is like, you can't even buy for three, four hundred bucks. So, I guess a lot of these startups, so at what ends up happening is that their robot, if you look at the bomb, it's at least $10,000 USD, if not more, yeah. So, the one that we're doing right now, I think the bomb is like less than 500 bucks. So, at least it's really like a toy. It's literally just, I think of it like an RC car except that, you know, you put a small computer, right now it just rest verify. We started with Justin Nano, but like I said, there's no more Justin Nano so we have to switch the rest verify. And to be frank, initially we wanted to run some like computer vision and self-driving thing. That was when we are still doing it as a hobby, but the more we think about it, the more we realize actually, it's almost like, it's a little bit contrary. Like I said, because most startups in this field, they try to do the deep learning thing. We are taking almost the opposite approach. What we are thinking is, it's almost, there's no point doing self-driving unless you have a tremendous amount of recorded data, kind of like what Tesla has done. And our view is that you probably need about one to 10 billion miles recorded before you should even attempt to, you know, kind of like what, what my brother does at Tesla, right? You know, like, so, so the question is how to break that GNA, how to get to that one to 10 billion miles recorded. So we thought maybe the way to do it is switch the other way, just make this robot damn cheap, right? Almost as cheap as a vacuum cleaner. So that perhaps, you know, small restaurant can easily afford a few, or maybe the blue ocean idea will be, maybe every household can afford one. And, and on the, but that's just a, just a device, right? But you, instead of relying sort of self-driving on AI, which to be frank right now, like no one in the world can do like what they call a tier four autonomy. You know, and most they can do is maybe, I mean, they claim tier four, but frankly they still need a human to kind of jog out all the way, right? So in our mind, if you need a human there, jog out all the way, might as well just have the human literally drive the robot all the way. So instead of thinking of this whole thing as the AI problem, we are actually thinking of it as a networking problem, first and foremost. So how do we solve 4G, you know, how do we solve for like, like WebRTC, right? Basically, I think of like Zoom, right? You know, you know, two-way traffic, you know, video and then you just plus another data channel. I mean, I'm making things very simplistic, but essentially something like that. And so we started doing a lot of testing. And so, okay, maybe I'll just show you, this is actually from my perspective, this is the older UI. We have a newer, like more, like almost like game like UI now, but effectively it's the same thing. So you can see a bunch, I mean, it's very simple, front cam, back cam, some basic stats coming from a robot, battery level, the GPS data. But the thing I want to point out is probably a bit hard to see, is this a distance log, but this is so-called FBT, so this is actually so-called our token. So the general idea is that, I mean, for those of you who are into crypto right there, there's this play to earn thing, and more recently they move to earn things. So very, very similar idea. Basically, we give you token base on how much you drive, something like this, yeah? I'm simplifying a little bit, but effectively that's why it is. And so we were basically trying to take the learnings from Helium and see whether we can use this so-called tokenomics to hopefully just how Helium got almost a million people around the world to pay for 500 bucks to get that hotspot right. Likewise, we want people to spend, maybe a million people around the world to get, say, 500 bucks to buy one of these. And then, but we do have another player, we need to incentivize people to actually drive this thing, right? So that's where the so-called drive to earn tokens come in. So, So you think it's not self-driving, but self-driving? Not self-driving, yeah. I'm driving all the way, yeah. But actually the end goal is the same. We want to do self-driving ultimately, but we realize there's almost no point doing self-driving unless you have, in our view, like one billion miles, right? So to get to one billion miles, you probably need half a million to one billion of these robots. I mean, if you, let's say this robot, usual range in a year, maybe say about 1,000 kilometers, then you literally need a, probably one million will get you there, right? To get to about one billion miles. And the thing is, I don't know whether the self-driving on the road is necessary to, it's not exactly Apple to Apple comparison because what my brother told me was that, actually in his view, driving on a sidewalk is technically more challenging than driving on the road. Because driving on the road, you have traffic rules, right? So you can actually recognize a sign, they say stop, you have the speed, you can only turn left at this junction. So it's very regulated, so for traffic. Whereas in a sidewalk, even though your speed is so-called slow, but your edge cases are your norm. Basically it's just like the long tail is just very, very long. So by, in that sense, you need possibly even more data because you just need a lot more fringe cases too, before your deep learning start to really learn or the difference in hours. So in that sense, he thinks it's actually a tougher thing to do. And if you look at Tesla, they have not how many miles of, but definitely way more than 10 billion miles, right? So when I say 10 million miles, it's probably underestimating, right? But having said that, as an engineering problem, probably it's maybe perhaps it's easier to get to a pragmatic solution because in the cell driving, in the road case, if you fail, someone dies, you see? In this case, this thing only costs 400, 500 bucks. It's not carrying diamonds, at most it's carrying maybe lunch or whatnot, right? So, and because this thing is tiny and it moves quite slow, so it's not gonna create any damage. In fact, it's the other way, it's more like other people abuse it, kick it around, but people don't die, you see? So in a sense, perhaps the requirement for, you don't need a six sigma 99% accuracy, maybe you just need 99.9 essay, right? And you hopefully you amortize it, especially if the cost is low, if you have enough miles underneath, you amortize it over the life of this robot and you could have actually turned over. So anyway, that's the rough idea. What else do I wanna share? Okay, actually, camping here, I got actually two of my teammates here, camping here is one of our main hardware engineers. So you have a very technical question again. I'm just giving you the overall concept of what we're trying to do. So, yeah, so basically that's what we're trying to do. We do right now have a small team in China. So actually to what you guys were discussing about PCB, you know, I think we're just very thankful that we even have a small presence in China because it's just so much easier to get things done hardware-wise. You know, the turnaround time, getting components, getting a PCB, you know, it's literally, they don't even need to think about it, just send it and you know, even it's one, two pieces, like within two, three days they get it back, right? So, yeah, we do have a kind of like remote team, Spain, Taiwan, you know, yeah, different places but we do have a small team in China, yeah. So, but right now we're still very early. Really what we have now is we're very, very prototype, clearly not meant for like to be sold to consumers. So we're working towards like getting into production, which is still quite a long way to go. Along the way, we discover a lot of things regarding to logistics and whatnot, you know, if you want to like, like even ship this thing to overseas, right? Like, you know, the battery consideration and all that. Happy to go into more details where we found a lot more things recently, but. I would say, like, having done this for about half a year now, my personal feeling is that there are some cities where, let's say Singapore, where pockets in the city where you have decent 4G coverage. Like as you can see, right, when it works, it does work. Oh, by the way, I didn't mention just now that the video you saw on TikTok or Instagram, right? So this robot is actually driven by one of our teammates in Philippines. So, and that's one of the key thing that we wanted to test, right? Obviously, sometimes I drive. You know, obviously I'm in Singapore. The robot is in Singapore. But we actually, what we do every day, in fact, right now, like the guy is driving around a very basic answer, but he's actually based in Philippines. Yeah, so what we wanted to try is, can someone from literally really cross ocean, right? And then like maneuver this robot that's like miles away, right? And part of the consideration is also like, if we really, this were really to do delivery, like why would someone use this, let's say, first to say using a human to deliver, right? I guess the obvious reason is you can, even though it's not AI, right? You're replacing one human with one human. I think the thought that is, let's say for a place like Singapore, beyond certain point, right, is let's say, grab food or food panda, right, of the world. They almost have an inelastic labor situation. Like beyond certain point, every additional delivery guy they need to make, they need to pay more. So they actually don't have economy or scale. They will never have economy or scale, especially in first year cities. So my brother-in-law sometimes write for grab food, right? So he tells me, sometimes he get on a good hour, he get maybe 15 to 20 bucks, right? Sing dollars or if he does maybe two, three delivery in an hour. Obviously, if you get someone from say Philippines, Indonesia, you can get way lower wages. And it's not just, I guess the demo, it's not just, I guess geography as well. It could also be demographics. So I mentioned this guy in our team from Philippines. So he's actually a gamer. He actually, you guys know this game play to earn like X infinity. So he actually played that game to earn tokens. Initially, when I first help out, he didn't know that this is a crypto project. But you see to him, right? To him, he's just playing a video game. Yeah, so I don't know whether I have the room here, but actually what we want to do, right? It's actually to, the new UX for this is gonna be a lot more like a game. So that would be like when you earn the token, you're thinking, you know, like the coins and you know, and then you can imagine there will be leaderboards and all that. And so we want to turn it into almost like a very fun thing to do and almost like you're traveling the world through the eyes of a robot, right? So hopefully on the supply on the labor, we can get like tremendous amount of supply of labor. It could be, let's say in lunchtime in New York now, but it could be like midnight in like this part of the world and there's so many people available. And frankly, to do this job all you need is a decent propane. Yeah, you don't need huge propane. I mean, as you can see, we're running at pretty low resolution. This is running at around 480p, right? The back camera and then this is 720p, something like that. And that what we found is sufficient to drive. A little bit on the latency as well. Obviously this is not like competitive gaming. The issue right now, actually one of the main issue we are grappling with is the latency is not stable. Yeah, so it does fluctuate and that makes it very hard to drive. But in general what we found is that if the latency is roughly half a second or lower, that means 500 millisecond or lower. And when I say latency, I mean a round trip. That means from the moment I say press forward on my controller, to the moment the video come back to me and I can see the thing going straight, but that is to us the round trip, right? We realized at this speed, which is around five kilometers per hour, it's actually relatively safe to drive. Yeah. So again, five kilometers per hour is like a fast walk. So I think for delivery trips that are relatively short, let's say two to three kilometers, which frankly is actually mixed up about 50 to 60% of all delivery. I've seen some charts where the delivery for grab food and a lot of this like doorknob dash and all that. If you, as you can imagine, it's very one-sided, right? So most of the trips, well I would say at least 50% is actually less than five kilometers. I think 60% less and then the other four. So I mean, obviously this cannot do far distance, but I think for those 50, 60%, which is pretty substantial chunk of the delivery, I think robots like this, we're not talking about decade away, I think we're talking about years away, maybe a couple of years, yeah. And you don't need AI, like I said, you just, however you do need decent 4G and you can work, you don't need 5G, you just need 4G, yeah. Okay, yeah. On the activity, I think you work a lot but practice everything together. So in terms of video feedback, it's about 4G. Everything is 4G, yeah. Everything is 4G, there is control, video, working, everything, yeah. Because that's the only connection that we have on this guy. But there's not any message about like Lora. Oh, Lora, oh no, that one was Helium. So Helium was built on Lora when, so we don't use Lora right now. Yeah. You want to measure that? Yeah. Yeah. How do you get along with the work, start support for? Yeah, great question. So I think it's, if you think of it almost like a repeated game, right? I think what we want to do is let's say if someone else have driven at this spot, right? Then you basically, or at least a 4G spot, then it's almost like waste, you know? Like you share here, you know, there's a Matah here, you know? So the other fellow driver don't come in. So I think there has to be a lot of this kind of like community sharing. Sorry, sorry. I think maybe I've talked too long already. But yeah, I mean, I'm, I think maybe I should stop first, but happy to answer more after that. I think Michael will still be hanging around, right? Yeah, yeah, yeah. Feel free to take a look at the photo. So I think it's just a time to share one more picture. Yeah. Yeah. Thank you so much. Thank you, Michael. Cool. Yeah. Okay. So that's it. Michael. Okay. All right. I didn't expect this to be recorded. Also, I didn't expect Chinmei to give the like low-end talk, basically. So this is even lower end than Chinmei. I'm also not a hardware guy. I'm a software guy. But basically every few years I kind of have a phase where I think like microcontrollers are cool and I just go online and I buy a lot of things, random stuff on like Cy-Tron or something. And so these are things I bought about like in December last year. This maybe or this LEDs. This thing as well, which is relevant later. It's also great for video calls actually if you have it on the table, side effect. But yeah, this is for like soldering and like seeing stuff. That's why there's a magnifying glass. So anyway, one day I was having all my stuff on my work desk or one night actually and I look up the ceiling and I see these nice colors because I had one of these LED rings next to this magnifying glass. And I realized, well, this is basically a projector, right? It's something that I made slides and a lens that focuses it. So I thought maybe my daughter will like that, right? She was about half a year old at that time. And I thought maybe she likes colorful lights. Seems like a safe bet. So I built a little prototype in a cardboard box and you can see here, I'm also doing like, I have like a star shaped, I don't know what you wanna call that stencil or something there, a raspberry pipe pickle. And yeah, just the magnifying glass on top. The results were okay. So I ordered the cheapest magnifying glass you can get on Amazon or the cheapest that you can get in like two days because I'm very impatient. So this all happened like within a couple of days because I lose focus almost immediately once something like that comes along. I did some 3D design for my 3D printer, which I barely know how to do. I think I have some, no, no, no, no, no successes. So, but it's kind of after a while after a couple of attempts, it came out sort of okay. I think that's the result. It's technically a video. I'm not sure if it's hi, hi, can actually play. So that's the way to it. I brought it with me. I'll plug it in later, but it's probably too bright here that the ceiling is too dark. Also maybe don't pick it up because it's kind of like half of those little things here they broke off. So it's, it doesn't really, it's very rickety. It's also, it has like this nice round shape which is accidental. It's just what is it called in 3D printing warping or something. So it's very, it's really like minimum effort. It falls apart if you touch it, but it still works. So I'll plug it in later and you can have a look at it. Yeah. Maybe what's next, probably nothing because as I said, I usually lose interest in these things very quickly, but there are, so this is an eight by eight LED matrix. You can get I think 32 by 32 that are just a little bit bigger. The magnifying glass I bought also has two lenses. So I was experimenting a little bit with like focusing and things like that. That was interesting. So like a future iteration could maybe be adjustable in terms of like how big the protection area is, et cetera. Yeah, and higher resolution. I tried to do like pixel art with this or just little like eight by eight pixel graphics and you can sort of recognize them, but it doesn't really look nice. So I just reverted to this random, just random fade in, fade out, random colors basically. Any questions? Roland? How much power do you need? Sorry? How much power do you need? Not a lot. It's actually in, so I hooked it up directly to the Pipe Kiko power, which you're not supposed to do, but it works. Yeah, yeah, it didn't melt. It worked reliably. I think it's really, it's part of our evening ritual when we put our daughter to sleep for a while, the light is on. She doesn't really care, unfortunately. Like she never really appreciated it or anything. Maybe it's just a bit too young. Yeah, maybe I just need to put in more effort. So if anybody has ideas for like awesome, awesome things to like 3D print and build for kids, that would be great. Maybe you need some speaker. Yeah, so it was actually, that was one thing I wanted to do. We had a, is it called a mobile? Like a thing you hang over the crib that like moves around and things. So my original idea was to replace the motor because we had like a wind up one and it sounded really crappy. And it had like, you know, just like all analog stuff. So I wanted to build an electronic version with a servo, no, not a servo, a stepper motor and just playing like old chip tunes. I was looking at like the, it's got the soundtrack from Rocky and like Eye of a Tiger or something just to get like a nice strong kit. But then before I finish it, by the time I got the amp to work, like the amp circuit, before I got that to work, she was already too old for this and started like, like pulling it down and everything. So, okay. Yeah, exactly. Cool. Yeah, I shall plug it in and you can go about mixing and everything. I'm not sure if Jin made, do you do like an end moderation? Put it up and then we can leave it there. Yeah. The volume is a bullshit, but I think one. Yeah. But can I make one? It's on the chat. Oh, yeah. Wait, am I supposed to speak into the microphone? Yeah, I already ended the video. So I still do the very little part there. I guess you can contact me for the files. Cost is very cheap. It's like less than, probably less than $56 in total, including magnifying glass and the microchip and all that stuff. Little bit of 3D printing. I'm happy to share the files, but they're not very great. As I said, it kind of falls apart. All right. At today, so the rest of the evening, we can either look at some of the hardware or just hang around to chat. But before that, last call for announcements. Anybody has any announcements? I know Valentine wanted to talk about Hacker Space. So let me bring that up. Nope. What did I open? Something that I shouldn't be opening. How is it open collective, right? I can't find my mouse though. Oh, there it is. Yay. Yeah, fine. Valentine, I'm one of the people behind, we actually need to record this. Oh, I'm one of the people behind Hacker Space. And, you know, we've been around since 2009. This is our third venue and we're still working on it. We've always been working on it. And so this is more like a call for supporters because Hacker Space is a community funded space. And we used to be like mostly membership based, but I think it's not really sustainable. We spend like about 3,005 a month on rent and we don't actually have that much income. So, you know, if you have some money to chip in, we would actually like to get the community to support and run the space instead of just depending on members. And, you know, this is our open collective page which is like a fundraising page. And if you can, you know, if you're interested and you think that we're doing good work, I mean, hopefully. Or if you just want to have a say, then, you know, you can just chip in for like, you can either sign up for membership or just sign up for like a $16 supporter tier and hopefully like many people, like many hands make like work and we can get like maybe 300 people to support the space and so we don't have to worry about money at the time. So yeah, that's it. If you have any questions or you want me to like show you around the space, just let me know after this. All right, thanks. Thanks. Thank you once again for Hacker Space for hosting us. The website page is opencollective.com slash Hacker Space and Stream. All right, any other announcements? Anyone looking for jobs? Anyone want to talk about their events? Okay, cool. Well then, if you have a talk or a project, sorry, if you're doing a project or a hack or something that you want to share the next Hacker, let us know either on the meetup or the Facebook or in person. And if you can host Hacker for the next situation, so there's probably going to be next month around the same time a month later, let us know if you can host us a space about as big as this. Otherwise, we can always come back to Hacker. Can I have a question? Yeah, please do. So yeah, if anyone of you are interested to try Hacker the Port, we should open it up to the public soon. So what happens that you can just go to the website, sign up for the store and then all you need is just a decent broadband and a controller. Any kind of, most people will come in or most game controllers will do. There's an important to your laptop and it should work. And we will actually have someone please click all of the blocks over here. You guys can all, and you can even give it a go. And if you want, if you're interested, I think you can wear on the mutual social media to call off the server and all that. So if you follow that, I think maybe in a month's time you'll finally get that ready and then you'll get a chance to do that. For the boss, yeah, for the VHS, for the program, for the special power, right? Yeah, we can see it. Okay, all right. If that's no other announcements, then thank you for coming to Hackerware. Feel free to hang out and check out some of the stuff that people have brought. And yeah, yeah. Venetan will show us around Hacker space as well. And see you next time. Thank you for coming. Thank you.