Challenge: machine learning repeated typed codes/code patterns to create automatic snippets or conversations

it would be a challenge, but I never saw that done through machine learning/in an automatic way. Currently I am working on my emacs setups to 'manually' run programs, just to play to extract patterns. Using some statistics, we may be able to extract patterns at a certain/statistically useful pattern, and apply it -> make it a much automatically process then creating our own snippets by hands.

with sublime text 2: it's at an unique position to do so - with it's python plugin and modern interface.