Remove unnecessary comma

Author: kchristin22

_posts/2024-11-04-reverse-mode-autodiff-of-cuda-kernels-final.md

@@ -482,7 +482,7 @@ Now that's easy. And, thus, cool.
 
 ### Future work
 One could claim that this is the beginning of a never-ending story. There are numerous features of CUDA that could be supported in `Clad`, some of them being:
-* Shared memory: Shared memory can only be declared inside a kernel. Since, `Clad` transforms the original kernel into a device function, no declaration of shared memory can be present there. There are ongoing discussions on the need of the overload functions and the produced function's signature. 
+* Shared memory: Shared memory can only be declared inside a kernel. Since `Clad` transforms the original kernel into a device function, no declaration of shared memory can be present there. There are ongoing discussions on the need of the overload functions and the produced function's signature. 
 * Synchronization functions, like `__syncthreads()` and `cudaDeviceSynchronize()`
 * Other CUDA host functions
 * CUDA math functions