Here's the scarab I mentioned before - I scanned the original with the MDX-20 and cut a new one out of machinable wax. It was finished with a 1/16" ball-nose end mill, which did a surprisingly good job of reproducing small details. It's also a lot more resistant to breaking than the smaller, more expensive tools I'm learning to use. So far I've broken two 1/64" end mills, both in their first 60 seconds of operation.
Here's another experiment. It's a model of Mount St. Helens, based on USGS digital elevation models. I'm sure there must be an easier way to generate stereolithography files from DEMs, but for this test I had to make a VRML file and convert it to STL format using an evaluation copy of Rhino.
I still haven't had much opportunity to mill PCBs with the machine, but I'm still finding other useful things to do with it. Today I fabricated a replacement lens for my Maglite out of Lexan. I figure that saved me at least $2. Or rather, it would have if the original lens had actually been broken. But at least I've got a spare, just in case.