diff --git a/.DS_Store b/.DS_Store
new file mode 100644
index 00000000..5c1416eb
Binary files /dev/null and b/.DS_Store differ
diff --git a/maze_transformer/evaluation/plot_maze.py b/maze_transformer/evaluation/plot_maze.py
index 10ceff1e..cf76d1c3 100644
--- a/maze_transformer/evaluation/plot_maze.py
+++ b/maze_transformer/evaluation/plot_maze.py
@@ -6,10 +6,9 @@
 import matplotlib as mpl
 import matplotlib.pyplot as plt
 import numpy as np
-from jaxtyping import Float
+from jaxtyping import Array, Bool, Float
 from matplotlib.cm import ScalarMappable
 from matplotlib.colors import ListedColormap, Normalize
-from muutils.tensor_utils import NDArray
 
 from maze_transformer.generation.lattice_maze import Coord, CoordArray, LatticeMaze
 
@@ -225,8 +224,9 @@ def show(self, dpi: int = 100, title: str = "") -> None:
         self.plot(dpi=dpi, title=title)
         plt.show()
 
-    def _rowcol_to_coord(self, point: Coord) -> NDArray:
-        """Transform Point from MazeTransformer (row, column) notation to matplotlib default (x, y) notation where x is the horizontal axis."""
+    def _rowcol_to_coord(self, point: Coord) -> Float[Array, "2"]:
+        """Transform Point from MazeTransformer (row, column) notation to matplotlib default (x, y) notation where x
+        is the horizontal axis."""
         point = np.array([point[1], point[0]])
         return self.unit_length * (point + 0.5)
 
@@ -286,7 +286,7 @@ def _plot_maze(self) -> None:
 
             self.ax.imshow(img, cmap=cmap, vmin=-1, vmax=1)
 
-    def _lattice_maze_to_img(self) -> NDArray["row col", bool]:
+    def _lattice_maze_to_img(self) -> Bool[Array, "row col"]:
         """
         Build an image to visualise the maze.
         Each "unit" consists of a node and the right and lower adjacent wall/connection. Its area is ul * ul.
@@ -317,7 +317,7 @@ def _lattice_maze_to_img(self) -> NDArray["row col", bool]:
             connection_values = scaled_node_values
 
         # Create background image (all pixels set to -1, walls everywhere)
-        img: NDArray["row col", float] = -np.ones(
+        img: Float[Array, "row col"] = -np.ones(
             (
                 self.maze.grid_shape[0] * self.unit_length + 1,
                 self.maze.grid_shape[1] * self.unit_length + 1,
@@ -351,12 +351,13 @@ def _lattice_maze_to_img(self) -> NDArray["row col", bool]:
         return img
 
     def _plot_path(self, path_format: PathFormat) -> None:
-        p_transformed = np.array(
+        p_transformed: Float[Array, "coord 2"] = np.array(
             [self._rowcol_to_coord(coord) for coord in path_format.path]
         )
+
         if path_format.quiver_kwargs is not None:
-            x: NDArray = p_transformed[:, 0]
-            y: NDArray = p_transformed[:, 1]
+            x: Float[Array, "x"] = p_transformed[:, 0]
+            y: Float[Array, "y"] = p_transformed[:, 1]
             self.ax.quiver(
                 x[:-1],
                 y[:-1],
@@ -392,7 +393,7 @@ def _plot_path(self, path_format: PathFormat) -> None:
             ms=10,
         )
 
-    def as_ascii(self, start=None, end=None):
+    def as_ascii(self, start=None, end=None) -> str:
         """
         Returns an ASCII visualization of the maze.
         Courtesy of ChatGPT
diff --git a/maze_transformer/generation/constants.py b/maze_transformer/generation/constants.py
index cd5f197c..adf5aae6 100644
--- a/maze_transformer/generation/constants.py
+++ b/maze_transformer/generation/constants.py
@@ -1,9 +1,9 @@
 import numpy as np
-from muutils.tensor_utils import NDArray
+from jaxtyping import Array, Int, Int8
 
-Coord = NDArray["x y", np.int8]
+Coord = Int8[Array, "x y"]
 CoordTup = tuple[int, int]
-CoordArray = NDArray["coords", np.int8]
+CoordArray = Int8[Array, "coords"]
 
 SPECIAL_TOKENS: dict[str, str] = dict(
     adj_list_start="<ADJLIST_START>",
@@ -19,7 +19,7 @@
     padding="<PADDING>",
 )
 
-DIRECTIONS_MAP: NDArray["direction axes", int] = np.array(
+DIRECTIONS_MAP: Int[Array, "direction axes"] = np.array(
     [
         [0, 1],  # down
         [0, -1],  # up
@@ -29,7 +29,7 @@
 )
 
 
-NEIGHBORS_MASK: NDArray["coord point", int] = np.array(
+NEIGHBORS_MASK: Int[Array, "coord point"] = np.array(
     [
         [0, 1],  # down
         [0, -1],  # up
diff --git a/maze_transformer/generation/utils.py b/maze_transformer/generation/utils.py
index c87cb3e5..cc0c7490 100644
--- a/maze_transformer/generation/utils.py
+++ b/maze_transformer/generation/utils.py
@@ -1,12 +1,12 @@
 import math
 
 import numpy as np
-from muutils.tensor_utils import NDArray
+from jaxtyping import Array, Bool
 
 
 def bool_array_from_string(
     string: str, shape: list[int], true_symbol: str = "T"
-) -> NDArray:
+) -> Bool[Array, "..."]:
     """Transform a string into an ndarray of bools.
 
     Parameters
@@ -20,8 +20,8 @@ def bool_array_from_string(
 
     Returns
     -------
-    NDArray
-        A ndarray with dtype bool.
+    Bool[Array, "..."]
+        An ndarray array with dtype bool and an unknown shape.
 
     Examples
     --------
diff --git a/maze_transformer/training/tokenizer.py b/maze_transformer/training/tokenizer.py
index 9f08f61e..e7e79244 100644
--- a/maze_transformer/training/tokenizer.py
+++ b/maze_transformer/training/tokenizer.py
@@ -4,7 +4,7 @@
 from typing import TYPE_CHECKING, Union  # need Union as "a" | "b" doesn't work
 
 import torch
-from muutils.tensor_utils import ATensor, NDArray
+from muutils.tensor_utils import ATensor
 from transformers import PreTrainedTokenizer
 from transformers.tokenization_utils import BatchEncoding
 
@@ -140,7 +140,7 @@ def batch_decode(
 
     def to_ascii(
         self, sequence: list[int | str] | ATensor, start=None, end=None
-    ) -> NDArray:
+    ) -> str:
         # Sequence should be a single maze (not batch)
         if isinstance(sequence, list) and isinstance(sequence[0], str):
             str_sequence = sequence  # already decoded