A decade ago a number of semantic catalogs started appearing. These catalogs gave identifiers to things, assigned them categories and asserted facts about them. Dubbed knowledge graphs, the intent is to describe the world in a machine readable way. These catalogs have proved incredibly useful, allowing publishers to organize their content management systems, powering machines that can win game shows and allowing search engines to guide users by interpreting their queries as being about “things not strings.” While useful, these catalogs are semantically limited. The connections entities participate in are sparse, requiring human understanding when decoding relationships and categorical membership. Entities are frequently identified by lucky linguistic matches rather than constraints against semantic intent. If machines are to understand our world and react intelligently to requests about it, knowledge graphs need to grow beyond catalogs, encoding things which stretch the notion of “fact” and act as semantic APIs for the real world.