[brennonbrimhall]

The devil's in the details. Most of my favorite cases of programming are very much behind the scenes:

• 254's elegant usage of VEX bump switches. You can see them in Barrage's picture on their Aerial Assist page. I assume (based on their placement) that they were used to pulse their intakes when they are grabbing the balls for their famous auto.
• 254 and 1114 both had a strip of LEDs on their robot that lit up when the flywheel was up to speed for their drivers.
• Both 987 and 4334 have implemented their own engine to parse out an autonomous scripting language. 4334's engine for Gordian is open source and can be found on GitHub. Though I've never used Gordian, I can imagine that it cuts down on compile/reboot time when writing auto routines.*
• 33 had a unique catapult design that involved a lead screw that they used to vary their shot. I don't know if they ever did, but this gave them the ability to vary their shot distance on the fly. The only other teams that I know of that could change their shot distance on the fly were all flywheel based (or, per billylo's post, pneumatic).
• While not used, a sophomore on the team developed some OpenCV for ball detection, with the intention that it would be used in an autonomous mode that would pick up our partner's misses. The ball detection code was written, but the corresponding auto mode was never implemented due to our decision to allocate resources in a manner more consistent with our priority list. Who knows...maybe we'll finally have some time to realize it for some of our off-seasons.

*An aside: I'm curious to see whether or not they (Gordian, and other in-house scripting languages) are displaced with other solutions: the roboRIO is supposed to have Java 8, which should allow usage of Jython and/or Rhino (Java engines for Python and Javascript, respectively) for Java teams.