All posts by Eric Schweikardt

O frabjous day! I’m happy to report that Richard Siemens, our 1997 Siplace 80-S20 dual-head, conveyorized Pick & Place machine is fully operational! The significance of this milestone has everything to do with the fact that we bought it in September and have been debugging and repairing it ever since. Here’s us celebrating in our parking lot with a few bottles of Prosecco. Do you have a kit of Cubelets? Every single Cubelet contains a few tiny circuit boards, and most of these boards have quite a few fiddly little electronic components soldered on top: resistors, capacitors, microcontrollers, etc. If you’re holding a Cubelet in your hands, you might be surprised to know that the bulk of those components were placed onto the board manually, with tweezers, by one of our assembly elves. We’ve manually placed about 30,000 boards worth of components by hand! And that’s crazy, as most visitors to our factory have pointed out. No longer. We began shopping for a Pick & Place machine this Summer. We’ve spent a lot of time playing electronics with our friends at SparkFun and have been inspired by the creative ways that they manage to manufacture their products here in Boulder, CO. They’re currently making use of three P&P machines: two small Manncorp models and a brand-new fancy Mydata monster that they’ve written about recently. The Mydata machine is pretty incredible, but it’s also $170,000. Yeah. So we bought an $11,000 used Siemens machine that had been sitting in a shuttered factory in Mexico. Yes, we found sand inside. I’m way too excited about Richard’s current functionality to turn this into a rant about a certain shady industrial equipment dealer. The short version is that the machine, advertised as “tested and working,” needed a few thousand dollars worth of parts, some motor control boards, belts, hoses, all new software, and a few hundred hours worth of tinkering. Four months later, here we are. But look how fast! Fast indeed. Richard (we named him after our favorite aerobics instructor) is rated at 21,000 components per hour. 2 heads! 12 nozzles on each head! Nozzle garages! I never thought that I’d be so excited by assembly robots.

Somehow I ended up as a mentor for Haxlr8r, the Shenzhen-based hardware startup incubator.  I think they just wanted my name on the list — I haven’t actually mentored anyone or heard anything from the Haxlr8r people in two years other than an interview request from Zach Smith, who manages the whole thing.  I picked one question to answer, but since they never published it, I guess my answer was wrong.  Here it is anyway for posterity and your enjoyment. Zach:  Many have said that the hardware landscape is looking more like software with lower costs to entry, better prototyping tools, and faster turnaround times.  Do you think this is accurate?  What are the critical/important differences in your mind?  Eric:  No way.  Maybe on the same sort of evolutionary time scale upon which humans become more like birds whenever people who can jump really high have kids, but nowhere near close.  Software is often really easy to make and distribute, but hardware is hard. We’re getting close to having Tony Stark basements.  The prototyping shop at Modular Robotics is pretty awesome; we can make complex, high-resolution electromechanical thingamajigs really quickly.  We can make things in our shop that we couldn’t have made five years ago.  We can iterate on designs for tiny robots several times in a day!  But the engineers working in our fancy shop only create a few of anything, not thousands, and it always takes an obscene amount of money, time, and energy. The critical difference between making hardware and software is that we’ve got a pretty amazing replication and distribution mechanism in place for software.  Write a browser extension, upload it to SourceForge, and you’re done.  Hardware needs to be soldered and tested and packed in little boxes then placed on the brown truck.  It’s really not trivial.  Hardware has mass. Comparing something that has mass to something that doesn’t is a nice place for metaphors but is hard to digest.  It’s like comparing ideas to tires, rock ballads to doll house furniture.  But hardware doesn’t have to be like software!  If everyone could make and replicate any sort of hardware stuff we want, we’d be buried in grey electromechanical goo.  SourceForge has more than 300,000 simultaneous projects that are currently ready for you to download for free.  If we could make almost instant, almost infinite copies of everyone’s shitty toy robot kit, we’d be in trouble.

I’m super-excited to announce our launch of the Bluetooth Cubelet today.  It’s not just another Cubelet.  You can use it to pair your Cubelets robot with your phone — read sensor values from far away or remote control your mobile robot.  Or you can connect your Cubelets robot to your PC and reprogram any connected Cubelets using the Cubelets CODE web application and a simple API. This changes everything.  You can re-program a Think Cubelet to behave like a different Think Cubelet.  You can reprogram your Sense Cubelets to sense and then Invert the signal, removing the Inverse block from your robot.  You can roll up your sleeves and reprogram some Cubelets to do lots of the wacky things that have been suggested in the forums. We can’t wait to see what you build. We’ve been working on the Bluetooth Cubelet and its supporting software (Control and Code) for more than a year.  For us, that’s an eternity!  But this is not a simple project, and we wanted to get it right.  Massively parallel tangible real-time modular distributed robot programming.  You know, for kids!

Big day at Modular Robotics yesterday.  We were scrambling to make some last minute preparations for the evening’s Cubelets Hackathon when the Internet 2012 Bus Tour rolled up.  The red and blue bus just started a two week trip across the country to raise awareness for net neutrality.  It’s organized by Reddit, and is actually McCain’s old campaign bus.  It’s been re-wrapped, as you can see, and we assume cleaned out very well.  Anyway, Alexis and Erik were on board, along with about ten members of the press and various other interesting people from AgLocal, AdWeek, and the Internet Association. We played with Cubelets, walked around our factory, and talked a little about how crazy it is to think about manufacturing consumer electronics in the USA these days.  We ate little cheese Cubelets, drank coconut water, and had some fun with the Rally Fighter, Local Motors‘ short-run, semi-custom, community designed car. Why did the bus decide to stop at Modular Robotics?  It turns out that they just asked around for the coolest company in town and found us.  Awesome.  But there’s a deeper synergy at work here that might not be apparent at first glance. Cubelets are the building blocks of intelligent systems.  But we’re taking a huge departure from the way the field of artificial intelligence has worked for the last 60 years.  Normal artificial intelligence is “top down.”  We write programs for robots telling them to do one thing, then another, then wait for something, etc.  Instead of writing a big fat complicated program for a robot, Cubelets are “bottom up”.  Their behavior emerges from lots of little simple robots each doing their own thing.  The magic happens when they all get together and we see things like steering or intentional-looking behaviors like wall avoidance or alarm-sounding just sort of happen. This is kind of how the Internet works too: it’s a tremendously huge system that relies on many, many users and devices and programs to run, and it gives rise to all sorts of higher-level emergent elements (like bullying or social networking or SETI@Home) that couldn’t happen without all of the complexity at its base. The thing is, it’s really hard to make positive changes in a huge complex system from the top-down.  Yes, we’d all like artists to make a buck, and yes, we’d all like movies to keep getting made, but we can’t get there with ham-fisted approaches like SOPA or PIPA.  We can’t put the future of the internet into a few senators’ hands who don’t even have a clear idea of what it is or how it works. We need a more nuanced approach to solving problems in complex systems.  Instead of letting the RIAA dictate the rules, we need to look a little deeper and understand how and why patterns emerge in complexity and how even tiny changes can have ripple effects.  This is what Cubelets were designed for: to give kids a model to build their own complex systems and start to develop intuitions about how the world works.  To see that real solutions are not black or white, red or blue, but require research, critical thinking, and understanding. Well.  We had a lot of fun with the merry pranksters on the bus.  Thanks for coming!  Here’s my favorite tweet about the Bus Tour’s visit to Modular Robotics. Indeed! Credits: Jon Hiller took these great photos.  Thanks Jon!

I’m not a fan of the US Postal Service.  They’re basically bankrupt, they’re bloated, they’re always increasing prices, and they deliver so much junk to me that I feel guilty for cutting down the rainforest every time I check my PO Box.  They’re always losing my mail too.  But, they’re cheap.  A lot cheaper than UPS for international Cubelets shipments, and a little bit cheaper than UPS for domestic shipments.  We just added a couple of USPS shipping options to the site, so if you’re in Japan or Finland, we can now ship your Cubelets to you for $40 instead of $140.

I’m happy to announce that we just scored a big fat venture capital investment from our friends at the Foundry Group. We’re pretty excited — this deal will give us the cash to do so much more than we’re able to do now. For the last two years, we’ve been building Modular Robotics from an academic project into a toy company. A project like this, of course, requires cash. Normally, companies like us get a VC investment, selling part of the company in exchange for cash. But if you’ve talked with me about this in the past, you’ll know that I’ve been an adamant bootstrapper. I’ve seen all sorts of companies get pressured by investors to release crappy products, and I didn’t want that to happen to us. We had the freedom to bootstrap because our patient customers were willing to pre-order kits that they didn’t receive for months, and because the National Science Foundation supported us with us with three SBIR grants. Grant funding and pre-sales allowed us to grow to 18 people, to build and ship almost 20,000 Cubelets, and to carry on adamantly pulling ourselves up by our bootstraps. We didn’t need any business guys in pleated pants telling us what to do! But a couple of months ago, I changed my my mind. Why? Spreadsheets. With all of these employees and all of these tiny robots, my job has changed quite a bit. I spend a lot more time in front of a spreadsheet now than I spend in front of the laser cutter. And when I started to make projections, to see where we might be in one, two, five years, there was no question that a cash infusion now would let us scale up to make a tremendously bigger impact in the future than we could by bootstrapping. Bootstrapping can work, and it’s worked swimmingly for us so far. I mean, here we are, 18 people in a cool lab with a ping-pong table and a production line sending thousands and thousands of tiny robots all around the world! But we can do more. Scaling up now, building out our factory in Colorado, and hiring a few great people will let us get Cubelets out to a far greater number of kids than if we stuck to our previous path. Another reason for taking on this investment has to do with speed. Since we’re just a bunch of kids in a robot lab, we’re building our business based on agility. We think we can get cool new hardware built, tested, and onto store shelves before the big toy companies have any idea what’s going on. But lately, we’ve been doing a lot of waiting. Waiting for parts because we can only afford to order small quantities at a time. And waiting for tooling changes because we’re not first priority to some of our vendors. We realized that there was a better way. We went straight to the best VC firm in the country. Seriously: they wrote the book. The Foundry Group has had amazing success with tech companies and they also happen to invest in all of our friends like Orbotix, Sifteo, and MakerBot. We called managing director Brad Feld and he immediately understood what we were trying to accomplish. And so here we are! I’m not sure that anyone reading this really cares about our financial strategy; I’m pretty sure that our audience consists mostly of kids, educators, technologists, alpha-geeks, engineers, and artists. Not people who read techcrunch. I thought I’d write about this here, though, because it means that we’ll be able to make Cubelets (and you should see the other robots coming down the pipeline!) significantly faster and better than we even dreamed about yesterday.

I just gave a talk about tiny robots at the TEDx Front Range event. Excitement! Bright lights! As I began, my slides didn’t appear. There was confusion, audience banter. The AV guy was giving me one of these. So I proceeded, waved my arms around a lot, and talked for ten minutes about how I think tiny robots are going to help our kids become smarter than we are. I won’t lie: I prepared to give this presentation with a lot of imagery, and when it all of a sudden didn’t appear, I got stressed. Spontaneous challenge! But everything was fine: tiny robots speak for themselves. For those of you who were there, I’m sorry the images didn’t pop up; the talk would have been better if they had. For those of you who weren’t, here are the slides!

I’m happy to report that we’re busy assembling Cubelets and are finally shipping again. These are the new KT06 boxes! We’ve got a lot of pre-orders to work through — if you have a pending pre-order, expect an email in the next day or two with a new estimated shipping date.

Yesterday was pretty exciting at Modular Robotics. As you may know, we’ve soldered all of the Cubelets printed circuit boards to date using electric skillets. Thirty thousand tiny PCBs. This might shock an electrical engineer, but the method that Sparkfun published a few years ago works astonishingly well. Astonishingly well for a few thousand boards using a $20 piece of kitchen equipment, but it was time for us to upgrade. This enormous truck arrived first thing in the morning. We rented a forklift for the day to lift our new twelve foot long reflow oven off the truck and bring it into our shop. Matt drove. This beauty is an Electrovert Omniflo 5. We bought it used from a broker in Washington. It’s got 5 heating zones and is programmed with specific temperature curves to solder our Cubelets boards perfectly. Since the old skillet method worked so well, we weren’t really sure that we were going to need a real reflow oven, but it turned out that the boards that got burned or a little undercooked had a nefarious way of making their way further down the assembly process and requiring a lot of time to fix later. So we’re reflowing. 480 PCBs so far and they’re all beautiful! I never thought I’d be so excited about a piece of industrial equipment.