The article makes the point that the LOC metric is confounded by duplication:

Copy and paste programming leads to high LOC counts and poor design because it breeds duplication.

This problem is not insurmountable. Compression tools work by finding duplication and representing copies as (more concise) references to the original.* The size of the compressed version is an estimate of the real information content of the original, with copies counted at a significantly discounted rate. The compressed size of code could be a more robust measure of the work that went into it.

* Sometimes this is done explicitly, other times it's implicit