One question language for the data and for the blueprint. Load a graph, ask it what it knows, then ask it what it is. Every question here is standard SPARQL, running in your browser on an open-source engine, and would run unchanged on any standards-based triple store.
Turtle, the W3C text format for RDF. Paste your export from the Ontology Builder, open a .ttl file, or use the sample.
The facts: which things exist and what has been said about them. Click any thing in a result to see everything the graph knows about it.
The same language, pointed at the ontology itself. Classes and properties are triples too.
Terms that have been collected but not yet modelled, with their preferred and alternative labels.
The proof that context is a first-class citizen: one query reads a fact and the rule that permitted it.
Write any SPARQL. It runs as you type. Prefixes from the loaded file are added for you.