diff --git a/src/test_benchmark.py b/src/test_benchmark.py
index 97821642..cc5e7d07 100644
--- a/src/test_benchmark.py
+++ b/src/test_benchmark.py
@@ -11,7 +11,7 @@
 import pytest
 from iterpy.iter import Iter
 from timeseriesflattener.aggregators import Aggregator, MaxAggregator, MeanAggregator
-from timeseriesflattener.feature_specs.meta import LookDistance, ValueFrame
+from timeseriesflattener.feature_specs.meta import ValueFrame
 from timeseriesflattener.feature_specs.prediction_times import PredictionTimeFrame
 from timeseriesflattener.feature_specs.predictor import PredictorSpec
 from timeseriesflattener.flattener import Flattener
@@ -50,7 +50,7 @@ def _generate_benchmark_dataset(
     n_features: int,
     n_observations_per_pred_time: int,
     aggregations: Sequence[Literal["max", "mean"]],
-    lookbehinds: Sequence[LookDistance | tuple[LookDistance, LookDistance]],
+    lookbehinds: Sequence[dt.timedelta | tuple[dt.timedelta, dt.timedelta]],
 ) -> BenchmarkDataset:
     pred_time_df = PredictionTimeFrame(
         init_df=pl.LazyFrame(
diff --git a/src/timeseriesflattener/_intermediary_frames.py b/src/timeseriesflattener/_intermediary_frames.py
index 319f25d8..81d1aeb4 100644
--- a/src/timeseriesflattener/_intermediary_frames.py
+++ b/src/timeseriesflattener/_intermediary_frames.py
@@ -6,17 +6,11 @@
 import polars as pl
 
 from ._frame_validator import _validate_col_name_columns_exist
-from .feature_specs.default_column_names import (
-    default_prediction_time_uuid_col_name,
-    default_timestamp_col_name,
-)
 from .frame_utilities.anyframe_to_lazyframe import _anyframe_to_lazyframe
 
 if TYPE_CHECKING:
     from collections.abc import Sequence
 
-    from .feature_specs.meta import ValueType
-
 if TYPE_CHECKING:
     import datetime as dt
 
@@ -27,8 +21,8 @@ class TimeMaskedFrame:
 
     init_df: pl.LazyFrame
     value_col_names: Sequence[str]
-    timestamp_col_name: str = default_timestamp_col_name
-    prediction_time_uuid_col_name: str = default_prediction_time_uuid_col_name
+    timestamp_col_name: str = "timestamp"
+    prediction_time_uuid_col_name: str = "prediction_time_uuid"
     validate_cols_exist: bool = True
 
     def __post_init__(self):
@@ -47,12 +41,12 @@ def collect(self) -> pl.DataFrame:
 class AggregatedValueFrame:
     df: pl.LazyFrame
     value_col_name: str
-    prediction_time_uuid_col_name: str = default_prediction_time_uuid_col_name
+    prediction_time_uuid_col_name: str = "prediction_time_uuid"
 
     def __post_init__(self):
         _validate_col_name_columns_exist(obj=self)
 
-    def fill_nulls(self, fallback: ValueType) -> AggregatedValueFrame:
+    def fill_nulls(self, fallback: int | float | str | None) -> AggregatedValueFrame:
         filled = self.df.with_columns(
             pl.col(self.value_col_name)
             .fill_null(fallback)
@@ -76,7 +70,7 @@ class TimeDeltaFrame:
     df: pl.LazyFrame
     value_col_names: Sequence[str]
     value_timestamp_col_name: str
-    prediction_time_uuid_col_name: str = default_prediction_time_uuid_col_name
+    prediction_time_uuid_col_name: str = "prediction_time_uuid"
     timedelta_col_name: str = "time_from_prediction_to_value"
 
     def __post_init__(self):
diff --git a/src/timeseriesflattener/feature_specs/default_column_names.py b/src/timeseriesflattener/feature_specs/default_column_names.py
deleted file mode 100644
index ae387204..00000000
--- a/src/timeseriesflattener/feature_specs/default_column_names.py
+++ /dev/null
@@ -1,6 +0,0 @@
-from __future__ import annotations
-
-default_entity_id_col_name = "entity_id"
-default_prediction_time_uuid_col_name = "prediction_time_uuid"
-default_pred_time_col_name = "pred_timestamp"
-default_timestamp_col_name = "timestamp"
diff --git a/src/timeseriesflattener/feature_specs/meta.py b/src/timeseriesflattener/feature_specs/meta.py
index 5eb32012..964e9ccb 100644
--- a/src/timeseriesflattener/feature_specs/meta.py
+++ b/src/timeseriesflattener/feature_specs/meta.py
@@ -1,31 +1,15 @@
 from __future__ import annotations
 
 import datetime as dt
-from collections.abc import Sequence
 from dataclasses import InitVar, dataclass
-from typing import TYPE_CHECKING, Literal, Union
+from typing import Literal
 
 import pandas as pd
 import polars as pl
 
-from timeseriesflattener.feature_specs.default_column_names import default_entity_id_col_name
-
 from .._frame_validator import _validate_col_name_columns_exist
 from ..frame_utilities.anyframe_to_lazyframe import _anyframe_to_lazyframe
 
-if TYPE_CHECKING:
-    from typing_extensions import TypeAlias
-
-
-ValueType = Union[int, float, str, None]
-InitDF_T = Union[pl.LazyFrame, pl.DataFrame, pd.DataFrame]
-
-
-LookDistance = dt.timedelta
-
-
-LookDistances: TypeAlias = Sequence[Union[LookDistance, tuple[LookDistance, LookDistance]]]
-
 
 @dataclass
 class ValueFrame:
@@ -37,12 +21,14 @@ class ValueFrame:
         Additional columns containing the values of the time series. The name of the columns will be used for feature naming.
     """
 
-    init_df: InitVar[InitDF_T]
-    entity_id_col_name: str = default_entity_id_col_name
+    init_df: InitVar[pl.LazyFrame | pl.DataFrame | pd.DataFrame]
+    entity_id_col_name: str = "entity_id"
     value_timestamp_col_name: str = "timestamp"
     coerce_to_lazy: InitVar[bool] = True
 
-    def __post_init__(self, init_df: InitDF_T, coerce_to_lazy: bool):
+    def __post_init__(
+        self, init_df: pl.LazyFrame | pl.DataFrame | pd.DataFrame, coerce_to_lazy: bool
+    ):
         if coerce_to_lazy:
             self.df = _anyframe_to_lazyframe(init_df)
         else:
@@ -63,8 +49,8 @@ def collect(self) -> pl.DataFrame:
 
 @dataclass(frozen=True)
 class LookPeriod:
-    first: LookDistance
-    last: LookDistance
+    first: dt.timedelta
+    last: dt.timedelta
 
     def __post_init__(self):
         if self.first >= self.last:
@@ -74,11 +60,11 @@ def __post_init__(self):
 
 
 def _lookdistance_to_normalised_lookperiod(
-    lookdistance: LookDistance | tuple[LookDistance, LookDistance],
+    lookdistance: dt.timedelta | tuple[dt.timedelta, dt.timedelta],
     direction: Literal["ahead", "behind"],
 ) -> LookPeriod:
     is_ahead = direction == "ahead"
-    if isinstance(lookdistance, LookDistance):
+    if isinstance(lookdistance, dt.timedelta):
         return LookPeriod(
             first=dt.timedelta(days=0) if is_ahead else -lookdistance,
             last=lookdistance if is_ahead else dt.timedelta(0),
diff --git a/src/timeseriesflattener/feature_specs/outcome.py b/src/timeseriesflattener/feature_specs/outcome.py
index 77f67d26..37806aab 100644
--- a/src/timeseriesflattener/feature_specs/outcome.py
+++ b/src/timeseriesflattener/feature_specs/outcome.py
@@ -1,12 +1,13 @@
 from __future__ import annotations
 
+import datetime as dt
 from dataclasses import InitVar, dataclass
 from typing import TYPE_CHECKING
 
 import polars as pl
 
 from .._frame_validator import _validate_col_name_columns_exist
-from .meta import LookDistances, ValueFrame, ValueType, _lookdistance_to_normalised_lookperiod
+from .meta import ValueFrame, _lookdistance_to_normalised_lookperiod
 
 if TYPE_CHECKING:
     from collections.abc import Sequence
@@ -20,12 +21,14 @@ class OutcomeSpec:
     """Specification for an outcome. If your outcome is binary/boolean, you can use BooleanOutcomeSpec instead."""
 
     value_frame: ValueFrame
-    lookahead_distances: InitVar[LookDistances]
+    lookahead_distances: InitVar[Sequence[dt.timedelta | tuple[dt.timedelta, dt.timedelta]]]
     aggregators: Sequence[Aggregator]
-    fallback: ValueType
+    fallback: int | float | str | None
     column_prefix: str = "outc"
 
-    def __post_init__(self, lookahead_distances: LookDistances):
+    def __post_init__(
+        self, lookahead_distances: Sequence[dt.timedelta | tuple[dt.timedelta, dt.timedelta]]
+    ):
         self.normalised_lookperiod = [
             _lookdistance_to_normalised_lookperiod(lookdistance=lookdistance, direction="ahead")
             for lookdistance in lookahead_distances
@@ -47,7 +50,7 @@ class BooleanOutcomeSpec:
     """
 
     init_frame: InitVar[TimestampValueFrame]
-    lookahead_distances: LookDistances
+    lookahead_distances: Sequence[dt.timedelta | tuple[dt.timedelta, dt.timedelta]]
     aggregators: Sequence[Aggregator]
     output_name: str
     column_prefix: str = "outc"
diff --git a/src/timeseriesflattener/feature_specs/prediction_times.py b/src/timeseriesflattener/feature_specs/prediction_times.py
index 6605d411..77b2c6fa 100644
--- a/src/timeseriesflattener/feature_specs/prediction_times.py
+++ b/src/timeseriesflattener/feature_specs/prediction_times.py
@@ -3,21 +3,15 @@
 from dataclasses import InitVar, dataclass
 from typing import TYPE_CHECKING
 
+import pandas as pd
 import polars as pl
 
 from .._frame_validator import _validate_col_name_columns_exist
 from ..frame_utilities.anyframe_to_lazyframe import _anyframe_to_lazyframe
-from .default_column_names import (
-    default_entity_id_col_name,
-    default_pred_time_col_name,
-    default_prediction_time_uuid_col_name,
-)
 
 if TYPE_CHECKING:
     from collections.abc import Sequence
 
-    from .meta import InitDF_T
-
 
 @dataclass
 class PredictionTimeFrame:
@@ -28,13 +22,15 @@ class PredictionTimeFrame:
         timestamp_col_name: The name of the column containing the timestamps for when to make a prediction.
     """
 
-    init_df: InitVar[InitDF_T]
-    entity_id_col_name: str = default_entity_id_col_name
-    timestamp_col_name: str = default_pred_time_col_name
-    prediction_time_uuid_col_name: str = default_prediction_time_uuid_col_name
+    init_df: InitVar[pl.LazyFrame | pl.DataFrame | pd.DataFrame]
+    entity_id_col_name: str = "entity_id"
+    timestamp_col_name: str = "pred_timestamp"
+    prediction_time_uuid_col_name: str = "prediction_time_uuid"
     coerce_to_lazy: InitVar[bool] = True
 
-    def __post_init__(self, init_df: InitDF_T, coerce_to_lazy: bool):
+    def __post_init__(
+        self, init_df: pl.LazyFrame | pl.DataFrame | pd.DataFrame, coerce_to_lazy: bool
+    ):
         if coerce_to_lazy:
             self.df = _anyframe_to_lazyframe(init_df)
         else:
diff --git a/src/timeseriesflattener/feature_specs/predictor.py b/src/timeseriesflattener/feature_specs/predictor.py
index d6191e46..c5cef219 100644
--- a/src/timeseriesflattener/feature_specs/predictor.py
+++ b/src/timeseriesflattener/feature_specs/predictor.py
@@ -1,10 +1,11 @@
 from __future__ import annotations
 
+import datetime as dt
 from dataclasses import InitVar, dataclass
 from typing import TYPE_CHECKING
 
 from .._frame_validator import _validate_col_name_columns_exist
-from .meta import LookDistances, ValueFrame, ValueType, _lookdistance_to_normalised_lookperiod
+from .meta import ValueFrame, _lookdistance_to_normalised_lookperiod
 
 if TYPE_CHECKING:
     from collections.abc import Sequence
@@ -25,12 +26,14 @@ class PredictorSpec:
     """
 
     value_frame: ValueFrame
-    lookbehind_distances: InitVar[LookDistances]
+    lookbehind_distances: InitVar[Sequence[dt.timedelta | tuple[dt.timedelta, dt.timedelta]]]
     aggregators: Sequence[Aggregator]
-    fallback: ValueType
+    fallback: int | float | str | None
     column_prefix: str = "pred"
 
-    def __post_init__(self, lookbehind_distances: LookDistances):
+    def __post_init__(
+        self, lookbehind_distances: Sequence[dt.timedelta | tuple[dt.timedelta, dt.timedelta]]
+    ):
         self.normalised_lookperiod = [
             _lookdistance_to_normalised_lookperiod(lookdistance=lookdistance, direction="behind")
             for lookdistance in lookbehind_distances
diff --git a/src/timeseriesflattener/feature_specs/static.py b/src/timeseriesflattener/feature_specs/static.py
index 9dd6fa0d..4b6bc725 100644
--- a/src/timeseriesflattener/feature_specs/static.py
+++ b/src/timeseriesflattener/feature_specs/static.py
@@ -1,25 +1,21 @@
 from __future__ import annotations
 
 from dataclasses import InitVar, dataclass
-from typing import TYPE_CHECKING
 
+import pandas as pd
 import polars as pl
 
 from .._frame_validator import _validate_col_name_columns_exist
 from ..frame_utilities.anyframe_to_lazyframe import _anyframe_to_lazyframe
-from .default_column_names import default_entity_id_col_name
-
-if TYPE_CHECKING:
-    from .meta import InitDF_T, ValueType
 
 
 @dataclass
 class StaticFrame:
-    init_df: InitVar[InitDF_T]
+    init_df: InitVar[pl.LazyFrame | pl.DataFrame | pd.DataFrame]
 
-    entity_id_col_name: str = default_entity_id_col_name
+    entity_id_col_name: str = "entity_id"
 
-    def __post_init__(self, init_df: InitDF_T):
+    def __post_init__(self, init_df: pl.LazyFrame | pl.DataFrame | pd.DataFrame):
         self.df = _anyframe_to_lazyframe(init_df)
         _validate_col_name_columns_exist(obj=self)
         self.value_col_names = [col for col in self.df.columns if col != self.entity_id_col_name]
@@ -41,4 +37,4 @@ class StaticSpec:
 
     value_frame: StaticFrame
     column_prefix: str
-    fallback: ValueType
+    fallback: int | float | str | None
diff --git a/src/timeseriesflattener/feature_specs/timedelta.py b/src/timeseriesflattener/feature_specs/timedelta.py
index 1b6c54d5..ddf99aed 100644
--- a/src/timeseriesflattener/feature_specs/timedelta.py
+++ b/src/timeseriesflattener/feature_specs/timedelta.py
@@ -4,7 +4,7 @@
 from typing import TYPE_CHECKING, Literal
 
 from .._frame_validator import _validate_col_name_columns_exist
-from .meta import ValueFrame, ValueType
+from .meta import ValueFrame
 
 if TYPE_CHECKING:
     import polars as pl
@@ -15,7 +15,7 @@
 @dataclass
 class TimeDeltaSpec:
     init_frame: TimestampValueFrame
-    fallback: ValueType
+    fallback: int | float | str | None
     output_name: str
     column_prefix: str = "pred"
     time_format: Literal["seconds", "minutes", "hours", "days", "years"] = "days"
diff --git a/src/timeseriesflattener/feature_specs/timestamp_frame.py b/src/timeseriesflattener/feature_specs/timestamp_frame.py
index 6d956f9f..376ecba1 100644
--- a/src/timeseriesflattener/feature_specs/timestamp_frame.py
+++ b/src/timeseriesflattener/feature_specs/timestamp_frame.py
@@ -1,16 +1,12 @@
 from __future__ import annotations
 
 from dataclasses import InitVar, dataclass
-from typing import TYPE_CHECKING
 
+import pandas as pd
 import polars as pl
 
 from .._frame_validator import _validate_col_name_columns_exist
 from ..frame_utilities.anyframe_to_lazyframe import _anyframe_to_lazyframe
-from .default_column_names import default_entity_id_col_name
-
-if TYPE_CHECKING:
-    from .meta import InitDF_T
 
 
 @dataclass
@@ -22,11 +18,11 @@ class TimestampValueFrame:
         value_timestamp_col_name: The name of the column containing the timestamps. Must be a string, and the column's values must be datetimes.
     """
 
-    init_df: InitVar[InitDF_T]
-    entity_id_col_name: str = default_entity_id_col_name
+    init_df: InitVar[pl.LazyFrame | pl.DataFrame | pd.DataFrame]
+    entity_id_col_name: str = "entity_id"
     value_timestamp_col_name: str = "timestamp"
 
-    def __post_init__(self, init_df: InitDF_T):
+    def __post_init__(self, init_df: pl.LazyFrame | pl.DataFrame | pd.DataFrame):
         self.df = _anyframe_to_lazyframe(init_df)
         _validate_col_name_columns_exist(obj=self)
 
diff --git a/src/timeseriesflattener/frame_utilities/anyframe_to_lazyframe.py b/src/timeseriesflattener/frame_utilities/anyframe_to_lazyframe.py
index 3d7f9b61..99fa5cc5 100644
--- a/src/timeseriesflattener/frame_utilities/anyframe_to_lazyframe.py
+++ b/src/timeseriesflattener/frame_utilities/anyframe_to_lazyframe.py
@@ -1,15 +1,11 @@
 from __future__ import annotations
 
-from typing import TYPE_CHECKING
 
 import pandas as pd
 import polars as pl
 
-if TYPE_CHECKING:
-    from ..feature_specs.meta import InitDF_T
 
-
-def _anyframe_to_lazyframe(init_df: InitDF_T) -> pl.LazyFrame:
+def _anyframe_to_lazyframe(init_df: pl.LazyFrame | pl.DataFrame | pd.DataFrame) -> pl.LazyFrame:
     if isinstance(init_df, pl.LazyFrame):
         return init_df
     if isinstance(init_df, pl.DataFrame):
@@ -19,5 +15,5 @@ def _anyframe_to_lazyframe(init_df: InitDF_T) -> pl.LazyFrame:
     raise ValueError(f"Unsupported type: {type(init_df)}.")
 
 
-def _anyframe_to_eagerframe(init_df: InitDF_T) -> pl.DataFrame:
+def _anyframe_to_eagerframe(init_df: pl.LazyFrame | pl.DataFrame | pd.DataFrame) -> pl.DataFrame:
     return _anyframe_to_lazyframe(init_df).collect()
diff --git a/src/timeseriesflattener/process_spec.py b/src/timeseriesflattener/process_spec.py
index 3a230f8c..2f07f0b1 100644
--- a/src/timeseriesflattener/process_spec.py
+++ b/src/timeseriesflattener/process_spec.py
@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-from typing import TYPE_CHECKING, Union
+from typing import TYPE_CHECKING
 import datetime as dt
 
 from .feature_specs.static import StaticSpec
diff --git a/src/timeseriesflattener/spec_processors/temporal.py b/src/timeseriesflattener/spec_processors/temporal.py
index 679758a3..aeeb8709 100644
--- a/src/timeseriesflattener/spec_processors/temporal.py
+++ b/src/timeseriesflattener/spec_processors/temporal.py
@@ -8,17 +8,17 @@
 from iterpy.iter import Iter
 
 from .._intermediary_frames import ProcessedFrame, TimeDeltaFrame, TimeMaskedFrame
+from ..feature_specs.meta import ValueFrame
 from ..feature_specs.outcome import BooleanOutcomeSpec, OutcomeSpec
+from ..feature_specs.prediction_times import PredictionTimeFrame
 from ..feature_specs.predictor import PredictorSpec
 from ..frame_utilities._horisontally_concat import horizontally_concatenate_dfs
-from ..feature_specs.prediction_times import PredictionTimeFrame
-from ..feature_specs.meta import ValueFrame, InitDF_T
 
 if TYPE_CHECKING:
     from collections.abc import Sequence
 
     from ..aggregators import Aggregator
-    from ..feature_specs.meta import LookPeriod, ValueType
+    from ..feature_specs.meta import LookPeriod
 
 
 def _get_timedelta_frame(
@@ -105,7 +105,9 @@ def _mask_outside_lookperiod(
 
 
 def _aggregate_masked_frame(
-    masked_frame: TimeMaskedFrame, aggregators: Sequence[Aggregator], fallback: ValueType
+    masked_frame: TimeMaskedFrame,
+    aggregators: Sequence[Aggregator],
+    fallback: int | float | str | None,
 ) -> pl.LazyFrame:
     aggregator_expressions = [
         aggregator(value_col_name)
diff --git a/src/timeseriesflattener/v1/feature_specs/group_specs.py b/src/timeseriesflattener/v1/feature_specs/group_specs.py
index fcb1e2ca..e9ec383c 100644
--- a/src/timeseriesflattener/v1/feature_specs/group_specs.py
+++ b/src/timeseriesflattener/v1/feature_specs/group_specs.py
@@ -1,3 +1,4 @@
+from __future__ import annotations
 import itertools
 from dataclasses import dataclass
 from typing import Dict, List, Sequence, Tuple, Union
@@ -18,10 +19,10 @@ class NamedDataframe:
 
 
 class V1PGSProtocol(Protocol):
-    lookbehind_days: Sequence[Union[float, Tuple[float, float]]]
+    lookbehind_days: Sequence[float | Tuple[float, float]]
     named_dataframes: Sequence[NamedDataframe]
     aggregation_fns: Sequence[AggregationFunType]
-    fallback: Sequence[Union[int, float, str]]
+    fallback: Sequence[int | float | str]
     prefix: str = "pred"
 
 
@@ -109,8 +110,8 @@ def create_combinations(self) -> List[OutcomeSpec]:
 
 
 def create_feature_combinations_from_dict(
-    dictionary: Dict[str, Union[str, list]],
-) -> List[Dict[str, Union[str, float, int]]]:
+    dictionary: Dict[str, str | list],
+) -> List[Dict[str, str | float | int]]:
     """Create feature combinations from a dictionary of feature specifications.
     Only unpacks the top level of lists.
     Args:
diff --git a/src/timeseriesflattener/v1/feature_specs/single_specs.py b/src/timeseriesflattener/v1/feature_specs/single_specs.py
index 378b944d..bb1a8b2e 100644
--- a/src/timeseriesflattener/v1/feature_specs/single_specs.py
+++ b/src/timeseriesflattener/v1/feature_specs/single_specs.py
@@ -1,3 +1,4 @@
+from __future__ import annotations
 from dataclasses import dataclass
 from typing import Tuple, Union
 
@@ -21,7 +22,7 @@ def __post_init__(self):
 @dataclass(frozen=True)
 class CoercedFloats:
     lookperiod: LookPeriod
-    fallback: Union[float, int]
+    fallback: float | int
 
 
 def can_be_coerced_losslessly_to_int(value: float) -> bool:
@@ -79,7 +80,7 @@ def get_temporal_col_name(
     feature_base_name: str,
     lookperiod: LookPeriod,
     aggregation_fn: AggregationFunType,
-    fallback: Union[float, int],
+    fallback: float | int,
 ) -> str:
     """Get the column name for the temporal feature."""
     coerced = coerce_floats(lookperiod=lookperiod, fallback=fallback)
diff --git a/src/timeseriesflattener/v1/flattened_dataset.py b/src/timeseriesflattener/v1/flattened_dataset.py
index 6ce455db..05ebbbfa 100644
--- a/src/timeseriesflattener/v1/flattened_dataset.py
+++ b/src/timeseriesflattener/v1/flattened_dataset.py
@@ -724,7 +724,7 @@ def _check_that_spec_df_timestamp_col_is_correctly_formatted(self, spec: Tempora
                     f"{spec.feature_base_name}: Minimum timestamp is {min_timestamp} - perhaps ints were coerced to timestamps?"
                 )
 
-    def add_spec(self, spec: Union[Sequence[AnySpec], AnySpec]):
+    def add_spec(self, spec: Sequence[Union[AnySpec, AnySpec]]):
         """Add a specification to the flattened dataset.
 
         This adds it to a queue of unprocessed specs, which are not processed
diff --git a/src/timeseriesflattener/v1/logger.py b/src/timeseriesflattener/v1/logger.py
index d93a46ec..7e3088a5 100644
--- a/src/timeseriesflattener/v1/logger.py
+++ b/src/timeseriesflattener/v1/logger.py
@@ -2,7 +2,7 @@
 
 import logging
 from pathlib import Path
-from typing import Optional, Union
+from typing import Optional
 
 import coloredlogs
 
@@ -12,7 +12,7 @@
 def setup_logger(
     name: str,
     level: int = logging.DEBUG,
-    log_file_path: Optional[Union[str, Path]] = None,
+    log_file_path: Optional[str | Path] = None,
     fmt: str = "%(asctime)s [%(levelname)s] %(message)s",
 ) -> logging.Logger:
     """
diff --git a/src/timeseriesflattener/v1/misc_utils.py b/src/timeseriesflattener/v1/misc_utils.py
index 49c793d2..6d086d57 100644
--- a/src/timeseriesflattener/v1/misc_utils.py
+++ b/src/timeseriesflattener/v1/misc_utils.py
@@ -2,12 +2,13 @@
 
 utilities. If this file grows, consider splitting it up.
 """
+from __future__ import annotations
 
 import functools
 import logging
 import os
 from pathlib import Path
-from typing import Any, Callable, Dict, Hashable, List, Union
+from typing import Any, Callable, Dict, Hashable, List
 
 import catalogue
 import pandas as pd
@@ -84,7 +85,7 @@ def format_dict_for_printing(d: dict) -> str:
     )
 
 
-def load_dataset_from_file(file_path: Path, nrows: Union[int, None] = None) -> pd.DataFrame:
+def load_dataset_from_file(file_path: Path, nrows: int | None = None) -> pd.DataFrame:
     """Load dataset from file. Handles csv and parquet files based on suffix.
 
     Args:
diff --git a/src/timeseriesflattener/v1/testing/utils_for_testing.py b/src/timeseriesflattener/v1/testing/utils_for_testing.py
index 9392e636..95d9a849 100644
--- a/src/timeseriesflattener/v1/testing/utils_for_testing.py
+++ b/src/timeseriesflattener/v1/testing/utils_for_testing.py
@@ -1,6 +1,9 @@
 """Utilities for testing."""
+from __future__ import annotations
+
 from io import StringIO
-from typing import Any, List, Optional, Sequence, Union
+from typing import Any, List, Optional, Sequence
+
 
 import numpy as np
 import pandas as pd
@@ -82,7 +85,7 @@ def str_to_df(
     return df.loc[:, ~df.columns.str.contains("^Unnamed")]
 
 
-def _get_value_cols_based_on_spec(df: pd.DataFrame, spec: AnySpec) -> Union[str, List[str]]:
+def _get_value_cols_based_on_spec(df: pd.DataFrame, spec: AnySpec) -> str | List[str]:
     """Get value columns based on spec. Checks if multiple value columns are present."""
     feature_name = spec.feature_base_name
     value_cols = df.columns[df.columns.str.contains(feature_name)].tolist()
@@ -94,7 +97,7 @@ def _get_value_cols_based_on_spec(df: pd.DataFrame, spec: AnySpec) -> Union[str,
 
 
 def assert_flattened_data_as_expected(
-    prediction_times_df: Union[pd.DataFrame, str],
+    prediction_times_df: pd.DataFrame | str,
     output_spec: AnySpec,
     expected_df: Optional[pd.DataFrame] = None,
     expected_values: Optional[Sequence[Any]] = None,