Aside from the fact that it doesn't actually work at all (I'm using Safari 2.0.1/WebKit 412.7 with Mac OS X 10.4.2 Tiger), the implementation leaves much to be desired:

• There's a Perl script that has to talk to a separate OS X application, which it calls to the front on every request.
• The "integration" between the script and the application is nothing more than the script's assumption that, 10 seconds after asking the application to make a JPEG of the web page, it will have done so. If the page already loads after 1 second, the user waits around for 9 unnecessary seconds; if the page loads in 11, the user gets a broken image.
• The software leaves temporary files on your hard drive without attempting to clean them up.
• Paths must be edited in the Perl script before it will "work" at all.
• There are mistakes in the included HTML files, which will be apparent by observing your server's error log while attempting to access a page.

It's a nice idea, but really needs to be done quite differently.