These tools made it simpler to do things like tagging, user-defined data attributes, hierarchical data, flexible configuration data.
It was always possible to store much of this data using text fields and JSON.parse (or some such), but the new Postgres stuff makes the flexible datatypes queryable. And as built-in support for flexible data types comes online, you can throw away your custom serializers.
There's other Postgres datatypes that I haven't used - for geocoding/geosearching, network addresses, etc.
Links to the talk, slides, sample code, and some follow up articles: http://monkeyandcrow.com/blog/postgres_railsconf2013/
Throw in a JSON schema validator and we have full-blown validations for :hstore and :json!