I wonder what's stopping someone from producing an all in one device: Desktop/phone/hobbyist micro-controller
I tried this. The problem is the HUGE amount of cash you have to put up up-front to just get devboards and datasheets. Also, while BT/WiFi licensing is pretty straight-forward, GSM licensing is not - and don't forget the licensing costs for the chips...
Furthermore, the more functionality you want to embed on your board (I intended on a ethernet+poe and a powerline/power supply board, linked via MII to the CPU), the more layers you will need on the board.
And RF engineering competence is a must both for functionality and for EM radiation compliance, and this is not cheap too.
Altogether a too small market.
Chinese manufacturers do tend to have room to move on samples and MOQs, I think it's really important for someone considering attempting hardware on the smaller scale have someone in China assist with interacting with the manufacturers. A big issue is the large amount of miscommunication which always seem to happen when western companies try to deal with manufacturers directly.
Like, a preferably low-cost / revenue-split consulting company for nonprofits and startups which gets the customers in contact with the Chinese HW/SW vendors and provides engineering (gongkai ecosystem) and especially cultural advice (translations!).
Also I'd prefer something much thicker to have more battery and a more powerful CPU. Thicker means easier to dissipate heat since air can circulate more.
Something much thicker would kind of defeat the purpose of being portable.
I believe what we have today is limited more by physical constraints (i.e. I can´t see myself writing code on a 7/8 screen, or typing comfortably on a keyboard that small) than actual technology.
Already, both on the Surface or a small footprint laptop (say, 11' Macbook Air), for me the experience is way worse than what you can get on a slightly larger laptop (13').
Batteries always takes room. I'd prefer a thicker laptop with more autonomy. I dont understand thinner laptops, especially plastic ones, since plastic doesnt seem to disperse heat very well.
Couldn't you use GSM modules  in your product right-away?
Also let me connect my phone to a monitor and keyboard and use the thing as my development platform. All in one.
Maybe a modular breakout board connectible to the IO port would be a better startup idea.
I've never programmed for mobile devices but are the operating systems open enough to let it be possible to write a drivers for these things?
* Leave one in the basement or the garage logging data from some sensor
* Give one to a kid to hack on
* Play around with using it as a server, as a media player, as a mini-workstation
Yes, most phones are more powerful than a Raspberry Pi, but no one is using a RPi because it's the most powerful thing out there. It's cheap, hackable, and separate from our other devices, but can be networked with them.
A nice product package would be a low-end tablet with some general Linux distro pre-installed and a USB to GPIO device, priced around $75. That would be useful for when you want a little machine with a UI to control something. Anything like that on the market?
Traditionally USB transceivers were designed as either master or slave (host/peripheral), though with the rise of smart peripherals (embedded/mobile devices) came need for both in one device. Manufacturers could design product with 2 different USB transceivers or use one that is configurable to either master or slave.
If you examine any mini/micro USB port/plug you will find 5 pins, and 4 pins on a large (both A and B). That extra pin is actually a toggle for OTG transceiver mode. Of course manufacturers can decide to internally wire transceiver to any mode and expose only data/power pins, but that does not stop transceiver from being OTG. Exactly this is done in top Android tables from mentioned site - they have nVidia Tegra SoC which actually uses OTG transceivers.
I am not intimate with PC hardware (would expect master-only transceivers), but, as I have said before, all embedded SoCs I have laid my hands on (not too many) used either slave-only or OTG transceivers.
This one would be enough, but why stop.
A lot of space in RPi and similar boards is dedicated to I/O pins and jacks. Would you carry an 1" thick device (or even an 1.5 cm thick device) instead of your current phone?
RPi runs various flavors of Linux, RPi2 will also run Windows 10, and you can try to run an entirely different OS. Are you comfortable changing the OS of your phone? Is it easy?
Well, you see. A phone is a piece of your life-support infrastructure (think calling 911), while an RPi is an entirely different thing that most people want to keep separate.
Not really. The prices have dropped tremendously. I recently picked up the Alcatel C1 for just $9 at the local frys. Sure it is under powered compared to the flagship phones, but it is really capable enough for tinkering, it also has a touchscreen and comes installed with Andriod KitKat 4.2, which could also be rooted. 
The low end Lumia phones always go on sale and has amazing specs for what you pay. For example Lumia 635 (on sale $19-$49) has IPS display, 5mp camera, Quad core 1.2GHz and a good amount of sensors.  However the gripe with these is these are not as hackable as Andriod, so good for backup phone but not so good for DIY projects.
But they ARE usable for many uses RPi is, and they do have a working 3G and WiFi, and also a touch screen, a speaker and a battery
I wonder how much of the phone cost is the display, cellular chips, microphones, speakers, light sensors, accelerometers etc?
Fernly is the work-in-progress OS for Fernvale: https://github.com/xobs/fernly
Using the shield form factor in the guide made it relatively easy to hold all the pieces together. Using smaller components would cry out for a case to hold them together.
http://www.ubuntu.com/phone plus an optional, well-integrated dock would be interesting though.
I can't even physically imagine that device, any more than I can imagine a trailer/motorcycle/airplane. (Well, I mean I can - just slide your phone apart to expose not a keyboard, but GPIO pins, and have an ugly HDMI out on it that turns it into a "desktop" stick. If I stretch my mind I can think of exactly one really good use for these specifications: handing them off to design team on April 1st. Few things are as exciting as the maker movement, Apps and phone development, or as powerful as desktop computing. With this in mind, this December we will be announcing the Android Phone and PC Environment Stick with Hobbyist Integrated Telemetry. Positioned directly at the intersection of desktop computing, a sleek and powerful phone, and an undergraduate electrical engineering course, project APPESHIT will do more than you ever thought possible in a mobile desktop phone microcontroller...)
I've tried different operating systems, hooked it up to all sorts of peripherals, left it running and hosting a VPN server in a drawer for an entire year...all things I'd never want to do with my phone, and all valuable learning experiences. A big part of the appeal is that it's not a computer you're relying on, so it's okay to break it (and fix it at your leisure).
For me, the two things I really like about the RPi (and BBB) are that I can plug in a full size keyboard and display, and that I can write and edit programs on the thing itself. Or I can use SSH or Remote Desktop, but I'm still programming directly on the target device.
I need the ergonomics of a full sized human interface due to my advancing old age, but I just have a visceral dislike for cross programming environments, where you program in an IDE on one computer, and run your program on another. Maybe this goes back to my earliest experiences with computers, where you flipped a switch (or entered a user name) and were instantly greeted with the BASIC prompt. I put up with cross development for microcontrollers, but still prefer a direct route.
Also, I'm not afraid to kill the RPi, and I can start over from a completely blank slate if I want to. When I think that I've gotten something working, I often wipe the system and re-install everything to make sure I know what would be required if someone else wanted to use my program.
As for my "case," I just attach everything to a piece of scrap plywood with wood screws, in the original sense of the "breadboard."
Rather than attach a keyboard, I connect to the RPi via ssh from my laptop. Compiling on the RPi itself is painfully slow in my experience.
I suppose eventually I will ditch x86 and my "laptop" will be ARM too for whatever that is worth. Then I can just focus on ARM assembly.
I never write to the SD card except when I am writing a new image with dd which is not very often. The root filsystem is in RAM (tmpfs). I can remove the card after boot so the slot is free.
I run this way from RAM on x86 too, even when I have a HDD. I like a diskless environment removable external storage for long-term data storage.
The only time I worry about USB stick or SD card read/write speeds is when I am putting a new image on the media with dd.
I still wish there was something like a next generation Jornada on the market. A rugged palmtop with Ethernet, WiFi, and USB slots that boots from CF card.
But in 2015 it looks like a development board on scrap plywood with wood screws is the only choice.
.. and its getting a new revision:
This is exactly the device you describe - powerful, complete, can be used in Desktop or handheld (or embedded) modes, open source, open community, no walled gardens or other limitations placed on developers ..
Just need a usb breakout cable for analog i/o