Because generators take iterables and produce iterables, you can chain them, each doing one small job. Values flow through the whole chain one at a time:

def numbers(limit):
    for n in range(1, limit + 1):
        yield n

def squares(values):
    for v in values:
        yield v * v

def only_odd(values):
    for v in values:
        if v % 2 == 1:
            yield v

pipeline = only_odd(squares(numbers(10)))
print(list(pipeline))

Generator expressions make short stages one-liners:

words = ["apple", "Banana", "cherry", "date", "Elderberry"]
lowered = (w.lower() for w in words)
long_ones = (w for w in lowered if len(w) > 5)
print(list(long_ones))

Nothing is computed until list pulls values through, and no full intermediate lists are built. This is exactly how large data processing is done.