The motionCAPTCHA author seems to be hilariously clueless. Read through the comments on the site from a year ago. A bunch of people, including Aaron Swartz, point out that it misses the fundamental point of a CAPTCHA (being a task that's easier for a human to carry out than a program) and his response to each is some variation of "you need to read the README more closely", or "I'll have a more production version that uses a regular PHP fallback out in a month".
I've been checking in every once in a while to see if there was actually a "secure", "production" version deployed. Figured I could throw together a bot to defeat it in a few minutes and that would be good for a laugh. Doesn't look like there has actually been any development done in a year though.