Re

So I just wrote this hierarchical kernelized Boltzmann process in Prolog using ed on my iPhone. I can send you the RCS repository.

— ML Hipster (@ML_Hipster) July 19, 2012

The best I can do is: I once programmed Prolog, wrapped inside Tcl and Ruby, to generate a query against a C++ IR engine, which generated features for a boosted decision tree trained in R, to rank documents indexed with that segfault-prone Tcl/Prolog/C system wrapped in Java/Hadoop. Then I wrote a Javascript/Ruby web interface on top.

Switching between these languages is too much for one days’ work. Educational, but not the best way to do things.