the json support is fairly recent and doesn't support operating directly on the data yet (e.g. you have to create functional indexes through plv8js and you can't directly query on the document's content as you can with array, xml or hstore columns)
Another intepretation of zapov's comment is that it's a shame the NoSQL projects didn't try achieving their goals by building postgres extensions (or even adding new features).
A reasonable question. MongoDB could have focused their energies on really good JSON support in postgres and making postgres sharding easier.