Back on November 2, in an item headed The single best way to stop spam is ... Bayesian, I linked to Paul Graham's statistical approach to creating a spam-filtering system that actually works. It was such a good idea, I thought it would disappear forever. But maybe not. Yesterday there was a spam conference at MIT where, according to Scarlet Pruitt of the IDG News Service, here, programmer William Yerazunis showed it implemented in a program called MailFilter. The report says: "In tests Yerazunis performed, MailFilter was 99.915 percent accurate in identifying spam. 'I'm only 99.84 percent accurate at identifying spam, so this is much more accurate than I am,' Yerazunis quipped." For more info, start here. [The Mailfilter program at Sourceforge is a different one.] AP and ComputerWorld also reported the MIT conference. Speaker John Graham-Cumming, who features in the latter story, is a former contributor to Guardian Online. William Yerazunis calls himself Bill.