Can someone describe a little bit on what the problems are with the current read prediction implementation (if the problems still exist)? Thanks a lot in advance! Ying