This is big, but it could be huge with one more element.
Right now, if you build some kind of appliance with the ADK, your market is limited to Android users. You're making a device for which the UI can be hosted on a user's mobile phone. But why does it have to be a phone? A proper Android competitor to the iPod would be a start, but here's what I would see as the game changer:
A $49 Android device with just a gigabyte or two of storage, no 3G, no camera, and an older ARM CPU, that can simply become the dedicated screen for something. You don't have to be an Android user and plug your phone into your exercise bike; your bike (or anything else in your home or office) just has a screen and a touch UI.
For $49 you're probably not going to get enough horsepower to run the whole system unless you can get the volume large enough. OMAP3 or i.MX5 could do it, maybe some of the later Chumby models could work.
Bonus: Your bike/whatever will automatically sync with an optional Android app based on your account. Also, some of the sub $100 cheapo Chinese tablets are the hardware you describe.
At I/O, didn't they say that those music cube things were running Android? It certainly looks as if they're planning on moving in that direction pretty soon.
edit: a bit of technical background. We all know that most Android devices and almost all phones really only support USB slave. This is mostly electrical, their controller chips almost always support OTG or Host, but they simply do not provide +5v on the USB bus to power an external device.
Hence in order to talk USB with your phone you had the option of rooting your device and use OTG (if supported), or to make the accessory play Host. In the latter case the accessory needs to talk the ADB (Android Debug Bridge) protocol to the phone, which is precisely what MicroBridge did using exactly the same hardware as ADK (Arduino + max3421e USB host controller). The ADB protocol provides byte stream multiplexing over USB, meaning that your USB-enabled microcontroller can tell your phone to give you a bidirectional byte stream to TCP ports, running daemons like logcat, a shell, etc. To establish a connection with a running app, that app simply listens on a TCP port, and the MCU opens a byte stream to that port using ADB.
What ADK does differently however, is that it adds a new USB protocol to Android. It previously supported three configurations: USB storage device, tethering, and ADB. The new ADK protocol is much simpler than ADB and really just only offers a single pipe with just reading/writing functionality. This is much simpler to implement (although ADB wasn't exactly rocket science), but most of all it's much cleaner and doesn't require the user to enable debugging in the phone settings.
ADB implementations have sprung up recently for various microcontrollers, such as PropBridge http://robots-everywhere.com/re_wiki/index.php?title=PropBri... for the propeller (which also comes in a DIP package btw, really nice for hobbyists), and IOIO for the PIC24: http://www.sparkfun.com/products/10585. These people will support the new protocol as well, so I expect there will be quite a bit of hardware choice in the near future.
This is a pretty big move. In the past, developing accessories for phones has had a huge overhead for development, and weren't easily distributed as expensive licenses from manufacturers were required, or users had to root/jailbreak their device. Consider how Square changed the payments market with their simple hardware device. Now we have the tools to allow more people to begin experimenting.
It's a huge opening for hobbyists and startups to really get a great experience with making some hardware for Android. I'm really excited to see what people can come up with.
This is a nice, but I think it's a step in the wrong direction.
It would be much better if Google had simply mandated that starting with, say, Android 3.1, all devices had to support USB OTG (http://en.wikipedia.org/wiki/USB_On-The-Go) without needing to be rooted.
What I'd really like to see is host mode being supported by Google from the get go. Sven Killig managed to get that running on his Nexus One to great effect. http://sven.killig.de/android/N1/2.2/usb_host/
Instead, we've running TCP, on top of ADB, running on top of USB.
Thanks for the link. I had wondered how multiple devices would work. (Plus, it's nice to see that I'm not the only one who came to the same conclusion - just add support for USB OTG.)
This whole ADK thing reeks of hardware-done-by-software guys. (Which is just as bad as software-done-by-hardware guys.) (I also hesitate to use the word engineer in this case.)
The AT90USB micro-controller has the same AVR core as the Ardunio which they're leveraging, and with LUFA, does fine in OTG or USB slave mode, and there are PICs that do USB if that's more your thing.
Electrically speaking, there is very little difference between a USB slave and a USB master, but unfortunately that's not something that's correctable in software.
Android is running Linux (specifically the kernel), which has drivers for a number of things, including USB card readers, printers, numerous USB-serial converters including FTDI, and other, more esoteric hardware too. Why re-invent the wheel, or in this case, drivers?
As a software engineer I object to that! At least software engineers would have thought their design through, instead this this whole thing looks thrown together by a bunch of amateurs over the weekend. I think it's more a testament of the 'there I fixed it!' attitude that's the cause of so much Android-related WTFs in general.
As for LUFA, I'm working on a port of MicroBridge to AT90USB. I'll get working on it as soon as I receive the dev board.
Still need to crack the final step-- how does a hobbyist make an accessory and get it to mass production. Kickstarter is one way, but I'm eager for JIT Manufacturing... or something.
I've been interested in that concept for a few years--the idea of "craft consumer electronics": small run, customised to a particular niche that isn't viable at large volumes.
I'd be really interested in finding people who were interested in exploring this area. I've had some conceptual thoughts around solutions but nothing concrete.
The thing with Quirky is that they're still gatekeepers. I've got to go through some pretty significant game mechanics. What would be useful is self-serve small run manufacturing. Or a lendingtree.com model for manufacturing, where you have a prototype and manufacturers bid on the project.
This represents an intriguing intersection of two big tech trends.
I'm teaching an introductory Arduino workshop at OSCON (O'Reilly Open Source Convention) again this year in July. Last year at OSCON Google had a big Android workshop, I'm curious what impact this announcement will have on interest in the Arduino workshop.
You can buy the reference design, or get an Arduino+USB host shield. You can get an Arduino anywhere, I'm not sure if their firmware will fit on a smaller AVR like the 328p. You can get the USB shield from Oleg's site directly (circuitsathome.com).
Sparkfun sells a cheaper version of Oleg's shield, although it seems to be out of stock right now: http://www.sparkfun.com/products/9947. Note that this shield doesn't work out of the box with the mega 1280/2560 boards because its SPI pins are located elsewhere compared to the smaller 168/328p Arduino boards like the Deicimila. You can do a hardware hack if you really want, but long story short, just get Oleg's shield.
edit: I've had a look at the code, and I can confirm this will work fine with either Oleg's shield or the Sparkfun shield, on a standard Arduino. So price-wise I'd say about $45 plus shipping from SF when they get their shields back in stock.
The site it links to, however, is less than useful for English speakers. (I think the site might be down under the load? I've been getting intermittent errors.)
Right now, if you build some kind of appliance with the ADK, your market is limited to Android users. You're making a device for which the UI can be hosted on a user's mobile phone. But why does it have to be a phone? A proper Android competitor to the iPod would be a start, but here's what I would see as the game changer:
A $49 Android device with just a gigabyte or two of storage, no 3G, no camera, and an older ARM CPU, that can simply become the dedicated screen for something. You don't have to be an Android user and plug your phone into your exercise bike; your bike (or anything else in your home or office) just has a screen and a touch UI.