diff --git a/efemel/pipeline.py b/efemel/pipeline.py
index 34492d5..aea2baa 100644
--- a/efemel/pipeline.py
+++ b/efemel/pipeline.py
@@ -13,10 +13,10 @@
 from concurrent.futures import FIRST_COMPLETED
 from concurrent.futures import ThreadPoolExecutor
 from concurrent.futures import wait
-from itertools import chain
 from typing import Any
 from typing import Self
 from typing import TypeVar
+from typing import Union
 from typing import overload
 
 T = TypeVar("T")  # Type variable for the elements in the pipeline
@@ -40,7 +40,7 @@ class Pipeline[T]:
 
   generator: Generator[list[T], None, None]
 
-  def __init__(self, source: Iterable[T], chunk_size: int = 1000) -> None:
+  def __init__(self, source: Union[Iterable[T], "Pipeline[T]"], chunk_size: int = 1000) -> None:
     """
     Initialize a new Pipeline with the given data source.
 
@@ -49,11 +49,13 @@ def __init__(self, source: Iterable[T], chunk_size: int = 1000) -> None:
                 If source is another Pipeline, it will be efficiently composed.
         chunk_size: Number of elements per chunk (default: 1000)
     """
-    if isinstance(source, Pipeline):
-      # If source is another Pipeline, use its generator directly to avoid double-chunking
-      self.generator = source.generator
-    else:
-      self.generator = self._chunked(source, chunk_size)
+    match source:
+      case Pipeline():
+        # If source is already a Pipeline, we can use its generator directly
+        self.generator = source.generator
+      case Iterable():
+        # If source is an iterable, we will chunk it
+        self.generator = self._chunked(source, chunk_size)
 
     self.chunk_size = chunk_size
 
@@ -97,38 +99,35 @@ def from_pipeline(cls, pipeline: "Pipeline[T]") -> "Pipeline[T]":
 
   def __iter__(self) -> Generator[T, None, None]:
     """Iterate over elements by flattening chunks."""
-    for chunk in self.generator:
-      yield from chunk
+    return (item for chunk in self.generator for item in chunk)
 
   def to_list(self) -> list[T]:
     """Convert the pipeline to a list by concatenating all chunks."""
-    result = []
-    for chunk in self.generator:
-      result.extend(chunk)
-    return result
+    return [item for chunk in self.generator for item in chunk]
 
   def first(self) -> T:
     """Get the first element from the pipeline."""
-    for chunk in self.generator:
-      if chunk:
-        return chunk[0]
-    raise StopIteration("Pipeline is empty")
+    item = next(self.generator, None)
+
+    if item is None:
+      raise StopIteration("Pipeline is empty")
+
+    return item.pop(0)
 
   def filter(self, predicate: Callable[[T], bool]) -> "Pipeline[T]":
     """Filter elements using a predicate, applied per chunk."""
 
-    def filter_chunk(chunk: list[T]) -> list[T]:
-      return [x for x in chunk if predicate(x)]
-
-    return Pipeline._from_chunks((filter_chunk(chunk) for chunk in self.generator), self.chunk_size)
+    return Pipeline._from_chunks(
+      ([x for x in chunk if predicate(x)] for chunk in self.generator),
+      self.chunk_size,
+    )
 
   def map(self, function: Callable[[T], U]) -> "Pipeline[U]":
     """Transform elements using a function, applied per chunk."""
-
-    def map_chunk(chunk: list[T]) -> list[U]:
-      return [function(x) for x in chunk]
-
-    return Pipeline._from_chunks((map_chunk(chunk) for chunk in self.generator), self.chunk_size)
+    return Pipeline._from_chunks(
+      ([function(x) for x in chunk] for chunk in self.generator),
+      self.chunk_size,
+    )
 
   def reduce(self, function: Callable[[U, T], U], initial: U) -> "Pipeline[U]":
     """Reduce elements to a single value using the given function."""
@@ -142,34 +141,26 @@ def tap(self, function: Callable[[T], Any]) -> Self:
     """Apply side effect to each element without modifying data."""
 
     def tap_chunk(chunk: list[T]) -> list[T]:
-      for item in chunk:
-        function(item)
-      return chunk
+      return [item for item in chunk if function(item) or True]
 
-    return Pipeline._from_chunks((tap_chunk(chunk) for chunk in self.generator), self.chunk_size)
+    return Pipeline._from_chunks(tap_chunk(chunk) for chunk in self.generator)
 
   def each(self, function: Callable[[T], Any]) -> None:
     """Apply function to each element (terminal operation)."""
-    for chunk in self.generator:
-      for item in chunk:
-        function(item)
+    deque((function(item) for chunk in self.generator for item in chunk), maxlen=0)
 
   def noop(self) -> None:
     """Consume the pipeline without any operation."""
     # Consume all elements in the pipeline without any operation
-    for _ in chain.from_iterable(self.generator):
-      continue
+    deque(self.generator, maxlen=0)
 
   def passthrough(self) -> Self:
     """Return the pipeline unchanged (identity operation)."""
     return self
 
-  def apply(self, *functions: Callable[[Self], "Pipeline[U]"]) -> "Pipeline[U]":
+  def apply(self, function: Callable[[Self], "Pipeline[U]"]) -> "Pipeline[U]":
     """Apply sequence of transformation functions."""
-    result: Pipeline[Any] = self
-    for function in functions:
-      result = function(result)
-    return result
+    return function(self)
 
   @overload
   def flatten(self: "Pipeline[list[U]]") -> "Pipeline[U]": ...
@@ -181,7 +172,7 @@ def flatten(self: "Pipeline[tuple[U, ...]]") -> "Pipeline[U]": ...
   def flatten(self: "Pipeline[set[U]]") -> "Pipeline[U]": ...
 
   def flatten(
-    self: "Pipeline[list[U]] | Pipeline[tuple[U, ...]] | Pipeline[set[U]]",
+    self: Union["Pipeline[list[U]]", "Pipeline[tuple[U, ...]]", "Pipeline[set[U]]"],
   ) -> "Pipeline[Any]":
     """Flatten iterable chunks into a single pipeline of elements.
 
@@ -201,9 +192,7 @@ def flatten(
 
     def flatten_generator() -> Generator[Any, None, None]:
       """Generator that yields individual flattened items."""
-      for chunk in self.generator:
-        for iterable in chunk:
-          yield from iterable
+      return (item for chunk in self.generator for iterable in chunk for item in iterable)
 
     # Re-chunk the flattened stream to maintain consistent chunk size
     return Pipeline._from_chunks(self._chunked(flatten_generator(), self.chunk_size), self.chunk_size)
@@ -312,6 +301,10 @@ def chain_generator():
       for pipeline in pipelines:
         yield from pipeline.generator
 
+    # chain preserves chunk structure exactly (just concatenates generators)
     # Use chunk_size from the first pipeline, or default if no pipelines
     chunk_size = pipelines[0].chunk_size if pipelines else 1000
-    return cls._from_chunks(chain_generator(), chunk_size)
+    new_pipeline = cls.__new__(cls)
+    new_pipeline.generator = chain_generator()
+    new_pipeline.chunk_size = chunk_size
+    return new_pipeline
diff --git a/tests/test_pipeline.py b/tests/test_pipeline.py
index cb6baf9..261d971 100644
--- a/tests/test_pipeline.py
+++ b/tests/test_pipeline.py
@@ -364,7 +364,7 @@ def test_passthrough(self):
     # Data should be unchanged
     assert result.to_list() == [1, 2, 3, 4, 5]
 
-  def test_apply_single_function(self):
+  def test_apply_function(self):
     """Test apply with single function."""
     pipeline = Pipeline([1, 2, 3, 4, 5])
 
@@ -374,27 +374,6 @@ def double_pipeline(p):
     result = pipeline.apply(double_pipeline)
     assert result.to_list() == [2, 4, 6, 8, 10]
 
-  def test_apply_multiple_functions(self):
-    """Test apply with multiple functions."""
-    pipeline = Pipeline([1, 2, 3, 4, 5])
-
-    def double_pipeline(p):
-      return p.map(lambda x: x * 2)
-
-    def filter_even(p):
-      return p.filter(lambda x: x % 2 == 0)
-
-    result = pipeline.apply(double_pipeline, filter_even)
-    assert result.to_list() == [2, 4, 6, 8, 10]
-
-  def test_apply_no_functions(self):
-    """Test apply with no functions."""
-    pipeline = Pipeline([1, 2, 3, 4, 5])
-
-    # Should return the same pipeline
-    result = pipeline.apply()
-    assert result.to_list() == [1, 2, 3, 4, 5]
-
   def test_flatten_basic(self):
     """Test basic flatten operation."""
     pipeline = Pipeline([[1, 2], [3, 4], [5]])