@@ -58,6 +58,12 @@ repos:
         exclude: '(.*pixi\.lock)|(\.git_archival\.txt)|(.*\.patch$)|(^cuda_core/cuda/core/_vendored/)|(^cuda_core/third_party/)'
         args: ["--fix"]
 
+      - id: check-cython-not-none
+        name: Check cdef-class-annotated def parameters carry 'not None'
+        entry: python ./toolshed/check_cython_not_none.py
+        language: python
+        files: ^cuda_core/cuda/.*\.pyx$
+
       - id: check-generated-file-seals
         name: Check generated-file seals
         entry: python ./toolshed/check_generated_file_seals.py
 
@@ -538,7 +538,12 @@ def module_names():
         "CUDA_CORE_BUILD_MAJOR": int(cuda_major),
         "CUDA_CORE_BINDINGS_FLOOR": tuple(_floor_for(cuda_major)),
     }
-    compiler_directives = {"embedsignature": True, "warn.deprecated.IF": False, "freethreading_compatible": True}
+    compiler_directives = {
+        "embedsignature": True,
+        "warn.deprecated.IF": False,
+        "freethreading_compatible": True,
+        "annotation_typing": False,
+    }
     _CythonOptions.warning_errors = True
     if COMPILE_FOR_COVERAGE:
         compiler_directives["linetrace"] = True
 
@@ -8,8 +8,6 @@ from collections.abc import Sequence
 from dataclasses import dataclass
 from typing import TYPE_CHECKING
 
-import cython
-
 from cuda.bindings cimport cydriver
 from cuda.core._device_resources cimport DeviceResources, SMResource, WorkqueueResource
 from cuda.core._device_resources import SMResource, WorkqueueResource
@@ -101,7 +99,6 @@ cdef class Context:
         Context_check_open(self)
         return DeviceResources._init_from_ctx(self._h_context, self._device_id)
 
-    @cython.annotation_typing(False)
     def create_stream(self, options: StreamOptions | None = None) -> Stream:
         """Create a new stream bound to this green context.
 
 
@@ -30,7 +30,6 @@ from cuda.core._rt cimport (
 )
 
 from cuda.core._stream import IsStreamType, Stream, StreamOptions
-from cuda.core._utils.clear_error_support import assert_type
 from cuda.core._utils.cuda_utils import (
     ComputeCapability,
     CUDAError,
@@ -1208,9 +1207,7 @@ class Device:
         return self._memory_resource
 
     @memory_resource.setter
-    def memory_resource(self, mr: MemoryResource) -> None:
-        from cuda.core._memory import MemoryResource
-        assert_type(mr, MemoryResource)
+    def memory_resource(self, MemoryResource mr not None: MemoryResource) -> None:
         self._memory_resource = mr
 
     @property
@@ -1247,7 +1244,7 @@ class Device:
     def __reduce__(self) -> tuple[object, ...]:
         return Device, (self.device_id,)
 
-    def set_current(self, ctx: Context | None = None) -> Context | None:
+    def set_current(self, Context ctx: Context | None = None) -> Context | None:
         """Set device to be used for GPU executions.
 
         Initializes CUDA and sets the calling thread to a valid CUDA
@@ -1291,8 +1288,6 @@ class Device:
         cdef Context prev_owned = None
 
         if ctx is not None:
-            # TODO: revisit once Context is cythonized
-            assert_type(ctx, Context)
             Context_check_open(ctx)
             if ctx._device_id != self._device_id:
                 # The CUDA context stack is per-thread, not per-Device-object,
 
@@ -475,7 +475,7 @@ cdef class SMResource:
 
     def split(
         self,
-        options: SMResourceOptions,
+        SMResourceOptions options not None: SMResourceOptions,
         *,
         bint dry_run=False
     ) -> tuple[list[SMResource], SMResource]:
@@ -587,7 +587,7 @@ cdef class WorkqueueResource:
                 "WorkqueueResource requires cuda.core to be built with CUDA 13.x bindings"
             )
 
-    def configure(self, options: WorkqueueResourceOptions) -> None:
+    def configure(self, WorkqueueResourceOptions options not None: WorkqueueResourceOptions) -> None:
         """Configure the workqueue resource in place.
 
         Parameters
 
@@ -168,7 +168,7 @@ cdef class Event:
     def __rsub__(self, other: object):
         return NotImplemented
 
-    def __sub__(self, other: Event) -> float:
+    def __sub__(self, Event other not None: Event) -> float:
         # return self - other (in milliseconds)
         Event_check_open(self)
         cdef Event other_event = Event_accept(other)
@@ -228,7 +228,7 @@ cdef class Event:
         return self._ipc_descriptor
 
     @classmethod
-    def from_ipc_descriptor(cls, ipc_descriptor: IPCEventDescriptor) -> Event:
+    def from_ipc_descriptor(cls, IPCEventDescriptor ipc_descriptor not None: IPCEventDescriptor) -> Event:
         """Import an event that was exported from another process.
 
         Parameters
@@ -377,7 +377,7 @@ cdef class IPCEventDescriptor:
         return IPCEventDescriptor._init, (self._reserved, self._is_blocking_sync)
 
 
-def _reduce_event(event: Event) -> tuple[object, ...]:
+def _reduce_event(Event event not None: Event) -> tuple[object, ...]:
     check_multiprocessing_start_method()
     return event.from_ipc_descriptor, (event.ipc_descriptor,)
 
 
@@ -223,7 +223,7 @@ cdef class GraphicsResource:
             return None
         return self._mapped_buffer
 
-    def map(self, *, stream: Stream) -> Buffer:
+    def map(self, *, Stream stream not None: Stream) -> Buffer:
         """Map this graphics resource for CUDA access.
 
         After mapping, a CUDA device pointer into the underlying graphics
 
@@ -29,8 +29,8 @@ __all__ = ['launch']
 
 def launch(
     stream: Stream | GraphBuilder | IsStreamType,
-    config: LaunchConfig,
-    kernel: Kernel,
+    LaunchConfig config not None: LaunchConfig,
+    Kernel kernel not None: Kernel,
     *kernel_args
 ) -> None:
     """Launches a :obj:`~_module.Kernel`
@@ -59,8 +59,7 @@ def launch(
     cdef ParamHolder ker_args = ParamHolder(kernel_args)
     cdef void** args_ptr = <void**><uintptr_t>(ker_args.ptr)
 
-    cdef Kernel ker = <Kernel>kernel
-    cdef cydriver.CUfunction func_handle = <cydriver.CUfunction>as_cu(ker._h_kernel)
+    cdef cydriver.CUfunction func_handle = <cydriver.CUfunction>as_cu(kernel._h_kernel)
 
     drv_cfg = conf._to_native_launch_config()
     drv_cfg.hStream = as_cu(s._h_stream)
 
@@ -96,7 +96,7 @@ cdef class _StridedLayout:
     @classmethod
     def dense_like(
         cls,
-        other : _StridedLayout,
+        _StridedLayout other not None : _StridedLayout,
         stride_order : str | tuple[int] = "K"
     ) -> _StridedLayout:
         """
 
@@ -404,8 +404,8 @@ cdef class Buffer:
     @classmethod
     def from_ipc_descriptor(
         cls, mr: DeviceMemoryResource | PinnedMemoryResource | VirtualMemoryResource,
-        ipc_descriptor: IPCBufferDescriptor,
-        *, stream: Stream
+        IPCBufferDescriptor ipc_descriptor not None: IPCBufferDescriptor,
+        *, Stream stream not None: Stream
     ) -> Buffer:
         """Import a buffer that was exported from another process.
 
@@ -505,7 +505,7 @@ cdef class Buffer:
         self.close()
         return False
 
-    def copy_to(self, dst: Buffer | None = None, *, stream: Stream | GraphBuilder,
+    def copy_to(self, Buffer dst: Buffer | None = None, *, stream: Stream | GraphBuilder,
                 options: CopyOptions | None = None) -> Buffer:
         """Copy from this buffer to the dst buffer asynchronously on the given stream.
 
@@ -555,7 +555,7 @@ cdef class Buffer:
                                  "buffer does not have a memory_resource)")
             dst = self._memory_resource.allocate(src_size, stream=s)
         else:
-            Buffer_check_open(<Buffer>dst)
+            Buffer_check_open(dst)
 
         cdef size_t dst_size = dst._size
         if dst_size != src_size:
@@ -566,7 +566,7 @@ cdef class Buffer:
             as_cu(dst._h_ptr), as_cu(self._h_ptr), src_size, s, options, "copy_to")
         return dst
 
-    def copy_from(self, src: Buffer, *, stream: Stream | GraphBuilder,
+    def copy_from(self, Buffer src not None: Buffer, *, stream: Stream | GraphBuilder,
                   options: CopyOptions | None = None) -> None:
         """Copy from the src buffer to this buffer asynchronously on the given stream.