The Fediverse (Lemmy/Mastodon/etc) is based on a following/subscribing model; each instance only “sees” what it’s users are currently following or subscribed to. This keeps storage and systems usage lower since each instance doesn’t need a complete copy of the entire Fediverse. This third party is more like a web crawler like Google, just crawling from instance to instance and saving the data. Hopefully in the future Lemmy could add something like this discovery feature, maybe something like Mastodon Relays, to aggregate community lists, but it would definitely put more strain on each instance.