Part of this power comes from the fact that Sequel's datasets use a functional style API.
With a functional style API, methods on datasets return a modified copy of the dataset, instead of modifying the datasets themselves.
This is powerful as it is thread-safe and allows for much more code reuse. You don't need to worry if another part of the code uses the dataset you are filtering.
Since filtering doesn't modify the dataset itself, but returns a copy, any code that uses the dataset itself is unaffected.