Ray already goes into most of the really interesting conclusions there on the page. I find myself wondering what a similar graph for me would look like; while I typically save big chunks of spam to train each new generation of heuristic filter, I don't have any unbroken chains anywhere near that long. Data like this is the source of so many of those "Hey, I wonder..." questions that sometimes end up producing the Really Good Ideas, though. Good on ya, Ray.
Posted by jbz at September 19, 2004 8:51 PM