I think it depends on your hardware and your purpose. I'll link to a couple of examples below.
I tried it on my iPhone 3GS and was extremely disappointed. It was difficult to know whether it was working/recording, and on the predictive lap timer display I didn't know what it was trying to tell me. It only records every other lap of video, which limits what it can be used for. It seemed dependent on a downloaded track profile to know where start/finish is for lap times; and the laps always start wrong (at least at Infineon, you can see it is about 2 seconds late). A buddy of mine used it on his iPhone 4 which has better GPS hardware and I think it worked better. It's supposed to also be much better when used with something like the Magellan GPS windshield mount. But the camera hardware still has the "rolling shutter" wobble effect, and if using the camera you can't see your lap times (until after the fact).
So for a "lap timer" you are better off with a lap timer. (I like the XT Racing Ultra-Lap for an IR timer.) And for good video you are better off with a cheap DV camcorder. Finally, the video angle when you put the phone on the windshield (which is necessary for GPS reception) is bad; the angle is much better if the camera is further back. You can combine the data with separate video (see example below) so that you have a good separate camera position and the GPS receiver on the windshield, but that's clearly more work to post-process the video.
On the other hand, for $20, what's not to like? Easy to experiment with. I might do one more experiment and then give up if it doesn't work better.
one example, showing the problemsusing a separate camera for video improves things