I have to say that I really like the trend toward "little" eval boards that plug into USB and sell for 'cheap.' I think TI started it with the 430F2013 USB stick, and the Cypress boards that have been mentioned are an excellent example. All pins available, usable but minimalistic programming interface, and easy to play with. If NXP put their LPC1102 on a DIP18 adapter with some decoupling and a serial bootloader, and sold it for $5, I'd be pretty happy. The price-point bar is pretty low, though. The LPCXPresso boards are only about $20, and so are for-profit boards like the "Teensy 3." $30 bucks for a mount-it-yourself bga adapter+chip isn't horrible if you really want to work with that chip, but it's not an enticing toy or impulse purchase, either.
I generally observe that if you want to catch engineers' imaginations and lure them to being interested in your products, you almost need to treat them like hobbyists. The number of occasions where a company or engineer will actually "evaluate" a wide range of chips/vendors using their "eval board" type offerings is very small. Usually you get a decision made based on someone's PRIOR or company legacy experiences (ranging from "I used an arduino in HS; let's use an AVR" to "our code base is 100k lines of code written for a 68000 that no one paid much attention to making portable. We're for damned sure stuck with a big-endian 32bit word chip and some new 68xxx compatible like CPU32 sounds a lot better than something more different."
(now, the BIG eval boards come in handy prototyping your software while your HW design is percolating through its cycle. The aforehinted 68k product had as much of its core as possible brought up on a 68331 eval board before the HW was done, and took about two days to get running on the actual HW when it came in.)