I feel like I just got the hang of it, so here let me try:

Imagine you live in a city (the internet). There are a bunch of good samaritans who have built massive houses (instances like kbin.social, mastodon.social, Lemmy.world, peertube etc.) that can host tons of people and also give you the construction plan for the houses, so you can choose to make your own kbin.bustanutella if you want to. Just because you're in a different house than your friends doesn't mean you can't communicate or interact with them. You don't even need to be a member of the same house to interact with them or any posts from their home. All that needs to happen is that both your house and your friends house need to agree to open their windows to each other, which is pretty standard as instances usually federate with each other freely unless an instance does something bad (spam, violating ToS, etc.)

So now, when your friend takes a shit on the floor of his house (makes a post), you get to see it and interact with it from the comfort of your own home. To interact with your friends, you would go to your own home and look through the window (login on your instance and search for the community URL or friend URL)

From all the use cases so far, it does seem like it is social media right now, but I don't see why it can't expand to other areas once adopted by enough people.

The Fediverse is a network of services. It is an agreement to communication between platforms.

Services can be groups of services with different themes and types of content.

Some people may create text-only services. They split into three groups of people, and each group hosts their own services that join the Fediverse. They then link each other and share their content and users. You can see and subscribe to and talk with users of the other servers.

Another group thinks text-only is stupid. They create a service that only hosts images. No text allowed. They join the Fediverse too.

Now both groups services are part of the Fediverse. They communicate and share content through the same kind of protocol, the same kind of agreements.

But does it make sense to mix text-only and picture-only services? Those that initially set out to create either may want a pure experience. They may present a split and separate experience to their users. But others may like both approaches, for an intentional, interesting form of content and discussion. They may create a service that combines the other two. In it, you can see and participate in both of the other kinds of services.

This and this.