Being able to parse without a lot of overhead and without allocations is quite interesting. E.g. when you process some massive json dump to just extract some properties (the Wikidata dumps come to mind).